Gondwana Large Igneous Provinces: plate reconstructions, volcanic basins and sill volumes

Abstract Gondwana was an enormous supertarrane. At its peak, it represented a landmass of about 100 × 106 km2 in size, corresponding to approximately 64% of all land areas today. Gondwana assembled in the Middle Cambrian, merged with Laurussia to form Pangea in the Carboniferous, and finally disintegrated with the separation of East and West Gondwana at about 170 Ma, and the separation of Africa and South America around 130 Ma. Here we have updated plate reconstructions from Gondwana history, with a special emphasis on the interactions between the continental crust of Gondwana and the mantle plumes resulting in Large Igneous Provinces (LIPs) at its surface. Moreover, we present an overview of the subvolcanic parts of the Gondwana LIPs (Kalkarindji, Central Atlantic Magmatic Province, Karoo and the Paraná–Etendeka) aimed at summarizing our current understanding of timings, scale and impact of these provinces. The Central Atlantic Magmatic Province (CAMP) reveals a conservative volume estimate of 700 000 km3 of subvolcanic intrusions, emplaced in the Brazilian sedimentary basins (58–66% of the total CAMP sill volume). The detailed evolution and melt-flux estimates for the CAMP and Gondwana-related LIPs are, however, poorly constrained, as they are not yet sufficiently explored with high-precision U–Pb geochronology.

The Gondwana supertarrane had a surface area of about 100 × 10 6 km 2 , and was assembled in Late Neoproterozoic and Cambrian times. It later amalgamated with Laurussia to form Pangea in the Late Carboniferous. Gondwana (or Gondwanaland) was first named by Medlicott & Blanford (1879), and included most of South America, Africa, Madagascar, India, Arabia, East Antarctica and Western Australia (Torsvik & Cocks 2013). Gondwana was affected by several Large Igneous Provinces (LIPs) from the Cambrian to its final break-up in the Cretaceous. These LIPs were likely to have been sourced from mantle plumes, and had major impacts on deep and shallow crustal rheology and on the Earth's climate. In general, the structure of the LIPs is two-fold, with (1) surface lava flows and pyroclastic deposits, and (2) a network of subvolcanic sills and dykes. The latter were commonly emplaced in sedimentary basins, and thus led to a wide range of effects including metamorphism, devolatilization, porosity reduction, and phreatic and phreatomagmatic eruptions.
The aim of this paper is to present plate reconstructions of the evolution of Gondwana, and to assess the timing and consequences of LIP formation. This is of great interest since Gondwana continental LIPs are associated with both major mass extinctions and rapid climatic changes. We stress that our contribution is not meant as a comprehensive review but as an update on key periods of the long Gondwana history (pertaining to the LIPs: Kalkarindji, Central Atlantic Magmatic Province, Karoo and the Paraná-Etendeka), and a useful summary of our current understanding. We put a special emphasis on the emplacement environments of the subvolcanic parts of the LIPs, in particular of the Central Atlantic Magmatic Province (CAMP) and the affected sedimentary basins, as this theme is poorly known outside the Brazilian geological community.

Plate reconstructions
We used GPlates (www.gplates.org; and GMAP (Torsvik & Smethurst 1999) for plate reconstructions and data analysis. Mesozoic reconstructions use a palaeomagnetic reference frame based on the zero Africa longitude method (Torsvik et al. 2012), whilst our Cambrian reconstruction is detailed in Torsvik et al. (2014). The lower mantle is dominated by two equatorial and antipodal regions of low seismic shear-wave velocities, referred to as the Large Low Shear-wave Velocity Provinces (LLSVPs), or using more friendly acronyms of Kevin Burke, TUZO (the LLSVP beneath Africa) and JASON (its Pacific counterpart). TUZO and JASON are prominent in all global shear-wave tomographic models, and most reconstructed Large Igneous Provinces (LIPs) and kimberlites over the past 300 myrand perhaps much longer (Torsvik et al. 2014)have erupted directly above their margins, termed the plume generation zones (PGZs: Burke et al. 2008). In order to relate deep-mantle processes to surface processes in a palaeomagnetic reference frame, we have counter-rotated the PGZs to account for true polar wander (TPW). Here we use the 0.9% slow contour in the tomographic shear-wave model (s10mean: Doubrovine et al. 2016) as the best approximation of the plume source region in the deep mantle. This is shown in Figures 1-3, where the location of the PGZ (thick red line) is corrected for TPW. In this way, we apply estimated TPW rotations (Torsvik et al. 2014) to the mantle to simultaneously visualize how the reconstructed continents and the underlying mantle structures would look like in the palaeogeographical reference frame (Torsvik & Cocks 2013) Geochronology Most of the LIPs described here (see Table 1) have been dated by high-precision zircon U-Pb chemical abrasion isotope dilution thermal ionization mass spectrometry (CA-ID-TIMS) geochronology, with the use of a common U-Pb tracer allowing direct interlaboratory data comparison (i.e. the EARTH-TIME tracer: Condon et al. 2015). The CA-ID-TIMS data from the Paraná part of the Paraná-Etendeka LIP (Janasi et al. 2011;Florisbal et al. 2014) and the data from Svensen et al. (2012) from the Karoo used in-house U-Pb tracers, but the tracer used in the latter study has been calibrated to the EARTH-TIME tracer (Svensen et al. 2015a;Corfu et al. 2016). There are no U-Pb data from the Etendeka part of the Paraná-Etendeka LIP, so only 40 Ar/ 39 Ar ages are reported (corrected to the latest standards). We have not included ion microprobe (SIMS/SHRIMP) and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) data as these methods do not yield ages with the required precision to evaluate potential correlations of short-lived geological events (e.g. LIP magmatism and mass extinction events). Furthermore, due to the relatively low level of precision of the data, potential Pb-loss is difficult to constrain, again leading to large uncertainties regarding the accuracy of ages. Similarly, baddeleyite ages are generally left out of this review unless they are combined with zircon and, hence, can be evaluated relative to zircon dataas these are also prone to Pb loss if not abraded (for a discussion of potential problems with baddeleyite ages, see, e.g., Corfu et al. 2016). All the zircon U-Pb ID-TIMS ages considered are reported, including tracer calibration uncertainties and excluding decay constants uncertainties, and are shown in Figure 4. For comparison, 40 Ar/ 39 Ar age ranges (recalculated to conform to the Fish Canyon sanidine age of Kuiper et al. 2008) are also included in Figure 4, as these ages represent the majority of analyses for several of the LIPs and has, in some cases, been used to argue for longevity of magmatic activity (e.g. Jourdan et al. 2005).

Basin-scale sill volume estimates
Total sill volumes from the sedimentary basins are notoriously difficult to estimate. Our approach is similar to that of Svensen et al. (2015b) from the Karoo Basin, using the aerial extent of the relevant basins and borehole-based sill thicknesses. Data from the literature (chiefly, outcrop, borehole and seismic data) were then scouted and used to estimate the total cumulative thickness of sills in the basins. Thicknesses of the sills are likely to vary across the basins, commonly with maxima at the centres; thus, minimum, maximum and mean thicknesses were used to calculate the total volume. We argue that the mean thickness is the most reasonable value to be used for estimating sill volumes in sedimentary basins, but minimum and maximum volumes are also reported in order to show the total range of plausible values. We summarize the literature-based sill volume estimates in Table 2, and our new CAMP compilation in Table 3, followed by a separate CAMP section in the Results.

The birth and demise of Gondwana
Unification of the many old cratons and terranes to form Gondwana began in the Late Neoproterozoic and continued into the Cambrian, but was largely over before the Middle Cambrian. The core of Gondwana included most of South America, Africa, Madagascar, India, Arabia, East Antarctica and Western Australia (Torsvik & Cocks 2013), and it reached a size of about 100 × 10 6 km 2 at its maximum (c. 64% of all land areas today). There were also many smaller units along its margins, such as Avalonia (including England), which had already drifted away during the Early Ordovician (with opening of the Rheic Ocean).
Gondwana merged with Laurussia (Laurentia, Baltica and Avalonia) in the Carboniferous at around 320 Ma to form Pangea. The disintegration of Gondwana had already started in the Early Ordovician but its demise after the Pangea assembly ( Fig. 1)

Constraining volumes: CAMP volcanism and sills in Brazil
Despite the growing amount of literature pertaining to the CAMP event, a thorough assessment of the preserved and original volumes of CAMP products is still lacking, with the only exception being a contribution by McHone (2003). In general, a total preerosional volume of 2.5 × 10 6 km 3 is addressed (e.g. Marzoli et al. 1999b). The total volume of magmatic products emplaced by the CAMP is within the same order of magnitude of those of other Large Igneous Provinces, but the general thickness (of outpoured material) is estimated as being smaller (Sebai et al. 1991;McHone 2003). Before we discuss the role of LIPs in the evolution and break-up of Gondwana, we would like to address the lack of constraint on the volumes associated with the CAMP event.
In general, intrusive and subvolcanic magmatism seems to constitute a large portion of the CAMP event. The only published estimates for volumes of CAMP products and relative degassing (McHone 2003) do not consider the existence of: (i) CAMP intrusions; (ii) Bolivian, Moroccan and Iberian CAMP occurrences (Knight et al. 2004;Bertrand et al. 2014;Callegaro et al. 2014); and (iii) an underestimate the extent of vast sills from Taoudenni-Hank-Reggane basins in Mali and Algeria (Verati et al. 2005;Charaf Chabou et al. 2010), from the Fouta Djalon Plateau (Deckart et al. 2005), and South America. Here we estimate the original volume of the CAMP starting from what is the presentday volume and surface of CAMP relicts. Particular emphasis is put on the sills and the intrusions, partly because the only available estimates to date are mainly focused on the lava piles, and partly because intrusive and subvolcanic bodies yield a high potential for degassing volatiles able to impact the global  ) and draped on the s10mean tomographic model of Doubrovine et al. (2016). The plume generation zone (PGZ) in this model corresponds to the 0.9% slow contour. LIPs with white-squared symbols are continental LIPs, whilst those with blue-squared symbols are oceanic plateaus (LIP numbers are ages in Ma: see Table 1). environment (e.g. Svensen et al. 2004Svensen et al. , 2009Ganino & Arndt 2009;Aarnes et al. 2010), making them significant in the broader picture of understanding the relationship between CAMP and the end-Triassic mass extinction Blackburn et al. 2013).
CAMP sills are often cropping out in restricted areas in sedimentary basins (outcrops cover c. 20% of the total), but borehole data from the basins generally intercept dolerites resting underneath most of the basin surface (cf., e.g., the Taoudenni Basin or the Brazilian basins: Milani & Zalán 1999;Verati et al. 2005). Boreholes and seismic data often reveal several sills stacked at different levels of the sedimentary pile (Fig. 5). We thus calculated the volume of CAMP sills by considering the extension of the basins that host them multiplied by the cumulative thickness of the doleritic sheets ( Fig. 6; Table 3). Depending on the structure of the basin, the thickness of the sills can vary from the centre to the edges, tapering towards both ends of the basin (e.g. within the Amazon Basin: Fig. 5) or as a regional trend (e.g. thickness waning east and west of the Solimões Basin: De Min et al. 2003), resulting in high uncertainty for the thickness estimation. Therefore, we calculated the minimum, average and maximum volumes by considering the variability related to the sill thickness (Table 3). The total estimated volume of CAMP products approaches 1 × 10 6 km 3 if the average thickness of sills and in-basin flows is considered. If a less conservative estimate is used, the total subvolcanic volume exceeds 1.6 × 10 6 km 3 .
It should be noted that these calculated volumes hinge on the extension of sedimentary basins and observation of the magmatic products preserved within them. Therefore, the better preservation of sills with respect to lava flows may lead to a disparity between intrusive and extrusive volumes that might not reflect the original proportion of CAMP products. While sills volume are well constrained through this method, lava-flow volumes should be taken as minimum estimates, since CAMP lava fields might have extended above all those regions presently cut by the dyke swarms (cf. McHone 2003). We decided to limit the calculations to basin areas because of the better-constrained information we have for these settings in the present: that is, being target to sedimentation, they are proven to be depocentres, capable of accommodating significant volumes of lava flows. Also, further potential CAMP occurrences are known to exist in remote areas of Africa and South America, but we here limit out assessment to the CAMP occurrences for which some (geochronological and/or geochemical) data have been published: that is, for which a CAMP affiliation was demonstrated. Therefore, we stress that these estimates are rather conservative and that future studies on the CAMP might result in larger volume estimates for this LIP.

The major LIPs of Gondwana
Gondwana is not only unique for having been the largest unit of continental crust on Earth for more than 200 myr before its amalgamation with Laurussia in the Carboniferous to form Pangea, but also for hosting abundant LIPs. Around 30 Phanerozoic LIPs are commonly listed in global compilations, and 19 of these (11 continental and seven oceanic LIPs, Table 1) -60% of all LIPsaffected Gondwanan continental lithosphere and its margins and oceanic lithosphere during Gondwana dispersal. Four case studies are discussed in this contribution, including the Cambrian Kalkarindji LIP in Western Australia, the end-Triassic CAMP that affected vast areas in North America, NW Africa and South America, the Early Jurassic Karoo-Ferrar LIP in South Africa and Antarctica (heralding the Jurassic     break-up of Gondwana), and, finally, the Paraná-Etendeka LIP, which affected large areas in South America and SW Africa (e.g. Brazil, Namibia and Angola), and assisted the opening of the South Atlantic from around 134 Ma.

The Kalkarindji Large Igneous Province
In the early Palaeozoic, Gondwana stretched from the South Pole to the equator (West Australia) and the Kalkarindji continental flood basalt province erupted at equatorial latitudes (Figs 1 & 3). Kalkarindji is a relatively poorly known LIP located in the interior of Western Australia. The name of the province was originally proposed by Glass & Phillips (2006) to encompass the Early Cambrian basalts of the Antrim Plateau in northern Australia, and a number of smaller basalt and dolerite outcrops in the interior of the continent. Recent overviews of the province can be found in Ernst (2014) and www.largeigneousprovinces.org/LOM The current extent of the Kalkarindji outcrops is about 55 000 km 2 , with an estimated initial areal extent of 400 000 km 2 and a volume of 0.15 × 10 6 km 3  and references therein). The areal extent of the entire Kalkarindji province is estimated to be at least 2.1 × 10 6 km 2 , but could be greater than 3 × 10 6 km 2 if igneous complexes in the Adelaide Fold Belt in South Australia are included as part of the province (Evins et al. 2009;Ernst 2014). The offshore extent of the province is currently unknown, but the Milliwindi dyke extends at least 70 km to the north of the Kimberly (Hanley & Wingate 2000). The volume of the erupted and associated shallow intrusive rocks is difficult to determine due to the scattered outcrops (see Fig. 7a), but is likely to be greater than 1.5 × 10 6 km 3 (Ernst 2014).
The igneous complexes of the Kalkarindji LIP consist dominantly of inflated, subaerial basalt flows and dolerite sheet intrusions. The basalts are dominantly low-Ti tholeiites and highly enriched in incompatible elements, suggesting continental contamination of the magmas (Glass & Phillips 2006;Ernst 2014). The most extensive and best-preserved basalt outcrops are located in the remote western part of the Antrim Plateau, reaching a thickness of at least 1.1 km. The lava flows are typically 20-60 m thick with massive interior and fractured or brecciated flow tops. Intraflow sedimentary units, including aeolian sandstones, are locally present. A basaltic agglomerate, the Blackfell Rockhole Member, has been described as an explosive phreatomagmatic tephra deposit but was recently reinterpreted as 25-40 m-thick rubbly pahoehoe lava flow ). In the Antrim Plateau, the extrusive pile was emplaced above Proterozoic clastic sedimentary basins, and is overlain by stromatolitic limestones of Late Cambrian age.
Subvolcanic complexes and the igneous plumbing systems are poorly exposed. An eruptive centre in the central part of the Antrim Plateau has been proposed by Glass & Phillips (2006). The Milliwindi dolerite dyke, a NW-trending, more than 200km-long intrusion in the NW Kimberly, has also been proposed as a potential feeder dyke for the extrusive basalts (Hanley & Wingate 2000).
An extensive Early Cambrian intrusive complex is present in the Officer Basin in central Australia (Jourdan et al. 2014). This Neoproterozoic intracratonic basin is more than 8 km deep with an areal extent of more than 300 000 km 2 . It is filled with marine carbonates, clastic sediments and evaporites. The structuring of the basin is mainly due to halokinesis. Dolerite sills are exposed in the NW Akubra-Boondawari area, and are penetrated by several petroleum exploration boreholes (Grey et al. 2005). Exploratory fieldwork in the Antrim Plateau region to the north revealed no presence of intrusive sills or hydrothermal vent complexes in the Proterozoic basins stratigraphically below the basalts.
A U-Pb zircon age of 510.7 ± 0.6 Ma obtained from the coarse-grained Milliwindi dolerite dyke is considered a minimum age for the onset of magmatic activity in the Kalkarindji LIP (Jourdan et al. 2014). This age overlaps the Early-Middle Cambrian boundary (510.0 ± 1.0 Ma: Landing et al. 1998;Harvey et al. 2011) and the associated extinction event (Fig. 4). The four 40 Ar/ 39 Ar ages that exist from the Kalkarindji LIP come from three flows and a sill, and range from 509.0 ± 2.6 to 511.9 ± 1.9 Ma, consistent with the U-Pb age (Glass & Phillips 2006;Evins et al. 2009;Jourdan et al. 2014).

The Central Atlantic Magmatic Province
Pangea was straddling the equator in the Late Triassic (Fig. 3), and the Central Atlantic Magmatic Province (CAMP) magmatic activity was located at equatorial-subtropical latitudes. As for many other LIPs, CAMP assisted a major plate tectonic reorganization that led to the opening of the central part of the Atlantic Ocean at around 195 Ma, and thus split the main area of the supercontinent into northern and southern Pangea, with the latter again often termed Gondwana. With a north-south extension exceeding 6000 km and laterally spread along more than 2500 km, CAMP is thus preserved (Marzoli et al. From the assessment presented here, we estimate that the total volume of CAMP products ranges within 1 × 10 6 -1.6 × 10 6 km 3 . It is clear from this analysis that the Brazilian basins host the majority of CAMP products, both in terms of sills (520 000 km 3 , c. 70% of the total volume of CAMP sills) and of lava flows (64 000 km 3 , c. 40% of the total volume estimated for basinal extrusive products). This assessment further highlights the fact that an important portion of the CAMP is constituted by intrusive and subvolcanic products, and that the majority of them are hosted by organic-rich sedimentary basins. Detailed studies of the magma-sediment interaction, coupled with high-precision U-Pb geochronology, will be the key for understanding the impact of CAMP on the global climate.
Only two high-precision zircon U-Pb CA-ID-TIMS studies have been published for the CAMP (Fig. 4) (Schoene et al. 2010;Blackburn et al. 2013). In addition, three air abrasion multigrain zircon and/or baddeleyite U-Pb ID-TIMS studies from two sills and one basalt in North America had previously been published (Dunning & Hodych 1990;Hodych & Dunning 1992;Schoene et al. 2006). Due to much larger uncertainties or unconstrained Pb-loss, these data are not included in  (Blackburn et al. 2013). Three sills are also dated from the North American part of CAMP, giving ages between 201.515 ± 0.062 and 200.916 ± 0.075 Ma, the latter being the youngest CAMP sample dated (Blackburn et al. 2013). One sill from Morocco, representing the only high-precision U-Pb date from the African part of the CAMP, is dated at 201.564 ± 0.075 Ma (Blackburn et al. 2013). Blackburn et al. (2013) estimated the end-Triassic extinction to have occurred at 201.564 ± 0.055 Ma, demonstrating synchronicity between early CAMP magmatism and extinction. Available 40 Ar/ 39 Ar ages show a relatively large range from c. 204 to c. 192 Ma, but with a significant majority clustering tightly around 201-202 Ma, also indistinguishable from the mean age of all CAMP magmatism ).

The Karoo-Ferrar Large Igneous Province and related basins
The bulk of Pangea was still rather intact by the Early Jurassic, with limited opening of the Central Atlantic. Karoo-Ferrar magmatic activity in southern Gondwana was centred on southerly latitudes (30-60°S: Fig. 3). The Karoo-Ferrar LIP covers a vast range of Gondwanaland areas. It is extensively preserved throughout southern Africa (e.g. South Africa, Namibia and Lesotho: du Toit 1920; Marsh et al. 1997;Neumann et al. 2011), in Tasmania andin Antarctica (e.g. Heimann et al. 1994;White et al. 2009). Invariably, the intrusive component of this LIP is emplaced within thick sedimentary sequences (e.g. the Karoo Supergroup in Africa and the Beacon Supergroup in Antarctica). In both the Antarctica sections and the Karoo sequences, the sill complexes can be mapped from the basement up through the main sedimentary succession (e.g. du Toit 1920; Marsh & Eales 1984;Galerne et al. 2008;Polteau et al. 2008;Jerram et al. 2010;Svensen et al. 2015b). By far, the most extensively studied parts of the Karoo-Ferrar LIP are in the Southern African portion and this provides us with an extensive geochronological framework expanded on below.
The Upper Carboniferous-Jurassic Karoo Supergroup in South Africa has a maximum cumulative thickness of 12 km and a preserved maximum thickness of 5.5 km (Tankard et al. 2009). The current area with Karoo sediments cropping out in South Africa is about 630 000 km 2 (Svensen et al. 2015b). The depositional environments range from marine to fluvial and, finally, aeolian (Catuneanu et al. 1998). The Karoo Basin is overlain by 1.3 km of volcanic rocks of the Drakensberg Group, consisting mainly of stacked basalt flows erupted into a continental and dry environment. The plumbing system of the flood basalts is a basin-scale sill complex consisting of sills and dykes (Marsh & Eales 1984;Chevallier & Woodford 1999). The thickest sill in the basin is about 220 m thick, most are in the 10-60 m range, and the sills were emplaced at about 182.6 Ma (Svensen et al. , 2015c. The composition of the sills is mainly tholeitic, with a few more evolved intrusions (andesitic) (Marsh & Eales 1984;Neumann et al. 2011).
Hundreds of breccia pipes and hydrothermal vent complexes are rooted in the contact aureoles in the Karoo Basin (Svensen et al. 2006(Svensen et al. , 2007. Alexander du Toit suggested that these degassing pipes formed due to igneous gas release, but recent research has stressed the importance of sediment-derived volatiles in their generation (e.g. Jamtveit et al. 2004;Svensen et al. 2007;Aarnes et al. 2010Aarnes et al. , 2012. The devolatilization involved both organic and inorganic reactions, leading to the formation of hightemperature minerals and lowering of the organic carbon content in shales (Aarnes et al. 2012;Svensen et al. 2015b). The hydrothermal vent complexes commonly crop out in the uppermost 400-500 m of the basin, and are associated with sills in the subsurface (Svensen et al. 2006(Svensen et al. , 2015b. In the upper parts of the basin stratigraphy (e.g. in the Elliot and Clarens formations), magma-water interactions led to the formation of phreatomagmatic complexes. One of the best-studied complexes, the Sterkspruit Complex, represents a >45 km 2 explosion crater filled with a variety of lavas, tuffs, sediment breccias, hyaloclastites and various other pyroclastic rocks (McClintock et al. 2008).
In Antarctica, the volcanic and subvolcanic rocks are scattered across vast areas along the Transantarctic Mountains and Queen Maud Land (e.g. Elliot & Fleming 2000;Elliot & Hanson 2001;McClintock & White 2006;White et al. 2009;Muirhead et al. 2014). Along the Transantarctic Mountains, a large sedimentary basin and its basement are exposed (e.g. Barrett et al. 1986;Storey et al. 2013). The sedimentary rocks are classified as the Beacon Supergroup and include Devonian-Triassic clastic sediments, mainly sandstones, also including Permian coal seams. Early Jurassic sedimentary rocks include sandstones intermixed and interbedded with tuffs and peperites (e.g. McClintock & White 2006;Muirhead et al. 2014). One of the most prominent sills in the Ferrar is the Peneplain sill, estimated to cover 19 000 km 2 with a thickness of 250 m (4750 km 3 ) (Gunn & Warren 1962;White et al. 2009). Other thick sills include the Basement sill and the Asgard and Mount Fleming sills, all of which attain thickness in the hundreds of metres and have a large aerial extent (e.g. Marsh 2004;Jerram et al. 2010).
There are three high-precision U-Pb ID-TIMS papers published from Karoo Sell et al. 2014;Burgess et al. 2015) and one extensive study from Ferrar (including 19 samples from three different areas of Antarctica and one sample from Tasmania: Burgess et al. 2015) (Fig. 4). The ages from Karoo obtained by Svensen et al. (2012) were mainly obtained on air-abraded zircons, with only a few chemically abraded zircons, but the study includes an extensive dataset on 14 samples with concordant zircons showing a large degree of consistency and, thus, most probably representing the true ages of the dated sills within uncertainty. The zircon ages from Sell et al. (2014) and Burgess et al. (2015) were obtained by CA-ID-TIMS. Only sills have been dated from Karoo and the ages range from 183.4 ± 0.4 to 182.7 ± 0.6 Ma Corfu et al. 2016), with the two most precise ages being 183.246 ± 0.066 and 183.014 ± 0.075 Ma (Sell et al. 2014;Burgess et al. 2015). In Ferrar, 20 samples from 10 lavas, eight sills and two intrusions have been dated by CA-ID-TIMS (Burgess et al. 2015). The lava ages range from 182.779 ± 0.066 to 182.430 ± 0.066 Ma, and the intrusive rocks range in age from 182.85 ± 0.35 to 182.540 ± 0.075 Ma (Burgess et al. 2015). Different astrochronological models exist for the onset of the Toarcian Ocean Anoxic Event (TOAE), one model placing the onset of the TOAE at 183.1 Ma (Boulila et al. 2014) and another at 182.75 Ma (Suan et al. 2008;Burgess et al. 2015), both anchored at the radiometrically determined Pliensbachian-Toarcian boundary age of 183.6 + 1.7/−1.1 (Pálfy et al. 2000;Sell et al. 2014). Whichever model is used, magmatism in Karoo overlaps the TOAE, while magmatism in Ferrar only overlaps with the TOAE using the model of Suan et al. (2008). Sills and lavas dated by 40 Ar/ 39 Ar show a relatively large spread (Fig. 4), but the majority of ages cluster between c. 182 and 186 Ma (Jourdan et al. 2008).

Paraná-Etendeka
By the Early Cretaceous, seafloor spreading in the Central Atlantic and the West Somali and Mozambique basins were well under way. The continents were spread from pole-to-pole and Paraná-Etendeka erupted at southerly subtropical latitudes (Fig. 3). The Cretaceous break-up of South America from Africa is a late-stage part of the Gondwana break-up and the Paraná-Etendeka LIP is the manifestation of this break-up; plume activity with volcanism started at 135-134 Ma, with large volumes of magma occurring in the first few million years up to, and including, the onset of break-up (e.g. Jerram & Widdowson 2005). This volcanism and the prevolcanic stratigraphy help to correlate the Paraná (Brazil) with the Etendeka (Namibia). A large volume of the preserved volcanic rocks are found on the Paraná side.
The volcanic stratigraphy of the Paraná-Etendeka, as for many of the LIPs, has been mapped on a gross scale using chemostratigraphic relationships (e.g. Paraná: Peate 1997; Etendeka: Marsh et al. 2001). Increasingly detailed field-based volcanological correlations of the lava sequences have also identified disconformities within lava sequences of the same chemostratigraphic type, as well as detailed understanding of the volcanic evolution of the province (e.g. Jerram et al. 1999;Jerram & Stollhofen 2002;Waichel et al. 2008). Also, key correlations with the base basalt sequences and the immediate underlying and interbedded stratigraphy (Jerram & Widdowson 2005;Petry et al. 2007;Waichel et al. 2012) have been made. Correlations across the South Atlantic are possible using these stratigraphic and chemical relationships between key units. Volumes calculated with help from these correlations highlight large-volume individual events, with some of the largest individual silicic eruptions adding up to several thousands of km 3 (Bryan et al. 2010). Such events are manifested as both eruptions and as intrusions with big sill and volcanic centre complexes (Jerram & Bryan 2015).
The Paraná-Etendeka has also been linked with climatic changes as recorded by carbon isotope excursions, specifically the mid-Valanginian Weissert Event (e.g. Erba et al. 2004;Aguirre-Urreta et al. 2015;Martinez et al. 2015). However, the scale and impact of this event is somewhat small compared to events such as the end-Permian. The volcanic basin of the Paraná-Etendeka contains significant aeoilan deposits (Mountney et al. 1998;Jerram et al. 2000), and great interaction of the volcanics with these sandstones is observed (Jerram & Stollhofen 2002;Petry et al. 2007). As such, there may have been a dampening down of the effects of gasses given off during this event compared to other LIPs as the interaction can often result in a net sequestration of gasses, such as CO 2 in diagenetic cements (Jones et al. 2016).
The geochronological picture from the Paraná-Etendeka is incomplete. Some U-Pb ID-TIMS age data include a combined zircon (multigrain fraction; air abraded) and baddeleyite (multigrain) weighted mean age from a dacitic volcanic rock (Janasi et al. 2011), and a five-zircon (multigrain; CA) upper intercept age from a composite dyke (Florisbal et al. 2014). No tracer calibration uncertainties were reported for these data but, as the uncertainties on the calculated ages are rather large, this is not a critical factor for comparison with other ages. The interpreted ages for the dacite and the dyke, respectively, are 134.3 ± 0.8 and 133.9 ± 0.7 Ma (Janasi et al. 2011;Florisbal et al. 2014), but due to the number few analyses and the clear presence of Pb loss in the analysed zircons, the accuracy of the ages cannot be considered as highly robust. Thiede & Vasconcelos (2010) reviewed 40 Ar/ 39 Ar ages from the Paraná and compiled the most robust ages. They also re-analysed some of the samples that deviated significantly from the mean age, and showed that all robust ages cluster tightly and define an interval of magmatism of less than c. 1.2 Ma, with a mean age of 134.6 Ma. Recent palaeomagnetic data suggest a longer lived magmatic range of c. 4 Ma, within a similar time frame (Dodd et al. 2015).
On the Etendeka (Namibian) side, there are only 40 Ar/ 39 Ar ages reported (e.g. Renne et al. 1996;Jerram et al. 1999;Marzoli et al. 1999a). They are stated as slightly younger (133-132 Ma) but when recalculated using the latest techniques for Ar/Ar calibration (Kuiper et al. 2008;Renne et al. 2010, see also Dodd et al. 2015), these come in at c. 134 Ma, overlapping with ages from the Paraná Basin (Fig. 4). Magmatic activity in the Paraná-Etendeka Magmatic Province appears to be slightly youngerbut overlapping within errorthan the onset of the mid-Valanginian Weissert Event (carbon isotope shift), dated at 135.22 ± 1.0 Ma (e.g. Aguirre-Urreta et al. 2015;Martinez et al. 2015). The question of synchronicity and matching between the geochronological data available for the two events remains open until better age constraints are attained.

LIPs, plates and the big picture
The dawn of the Phanerozoic is exceptional in many ways: most continents were located in the southern hemisphere, the atmospheric CO 2 concentration was, perhaps, 5-10 times the current level, and a global sea-level rise with largely warm surface seawater temperatures appear to have characterized the Cambrian and much of the Ordovician (Fig. 8). Abrupt changes in the atmospheric concentration of greenhouse gases have occurred episodically through Earth's history, and many of these climatic and environmental perturbations show a causal relationship with LIP eruptions, which have been sourced by plumes from the margins of two main thermochemical provinces in the deep mantle, TUZO and JASON. The Gondwana LIPs were all sourced from TUZO (Figs 2 & 3). During the Palaeozoic, most of the continents moved northwards, Pangea formed in the Late Carboniferous, and by the Late Triassic Pangea was centred around the equator and overlying the TUZO LLSVP (Fig. 3). CAMP magmatic activity was located to equatorial-subtropical latitudes, and contemporaneous kimberlites are found in southern Africa and North America. CAMP and kimberlites were sourced by deep plumes from the western margin of TUZO. The CAMP marks the initial break-up of the Pangaea supercontinent, relicts of its extensive tholeiitic magmatism are presently preserved in four continents along both sides of the Atlantic Ocean (Marzoli et al. 1999b), and CAMP contributed to one of the 'big five' biotic extinction events in the Phanerozoic.
By the Early Cretaceous, seafloor spreading in the Central Atlantic and the West Somali and Mozambique basins were well underway. The continents were spread from pole-to-pole and Paraná-Etendeka erupted at southerly subtropical latitudes (Fig. 3) with contemporaneous kimberlite activity in South Africa, Namibia and NW Africa. As for the end-Triassic and Early Jurassic, this magmatic activity was sourced by deep plumes along the western margin of TUZO. This is also exemplified in Figure 9: in this schematic cartoon we envisage that long-lived subduction along the South American (Andean) margin was the triggering mechanism for a deep plume that eventually rose through the mantle and led to catastrophic upper mantle melting, forming the Paraná-Etendeka LIP.
LIPs may have caused or contributed to four of the 'big five' biotic extinction events in the Phanerozoic: the end-Devonian (Yakutsk in Siberia), the end-Permian (Siberian Traps), the end-Triassic (Central Atlantic Magmatic Province, CAMP) and the end-Cretaceous (K-Pg: Deccan Traps). In addition, the Middle Permian extinction (Capitanian) was causally linked to the Emeishan LIP (e.g. Bond et al. 2010;Jerram et al. 2016b), but formed outside Gondwana and is thus not included in this work. The K-Pg mass extinction event, however, is unique because it is coincident with the Chicxulub bolide impact. The oldest Phanerozoic extinction, and the third in importance, at the end of the Ordovician (Hirnantian) is not linked with any known LIP, even though widespread explosive volcanism is suggested to have played a role in the global cooling (Buggisch et al. 2010).

The importance of subvolcanic intrusions and sill volumes
The timing of environmental changes and LIPs is summarized in Figure 8. Several processes are suggested as links between LIPs and environmental changes, including: (1) lava degassing of mantle volatiles, with or without a contribution from recycled continental crust; (2) degassing of volatiles derived from the sill-related contact aureoles of heated sedimentary rocks; and (3) degassing from sills and lavas contaminated by volatiles derived from the intruded sedimentary rocks or other crustal rocks (e.g. Svensen et al. 2004;Ganino & Arndt 2009;Sobolev et al. 2011). The actual extinction mechanisms are debated, and the suggestions include extreme temperatures, oceanic anoxia and pH reduction, sulphuric acid poisoning, and ozone layer destruction followed by extreme UV-B radiation (e.g. Visscher et al. 2004;Robock 2005;Bacon et al. 2013;Elliott-Kingston et al. 2014).
The importance of volcanic basins for the evolution of LIPs and the relationship to environmental changes has been stressed in recent years. During the formation of LIPs, magma is commonly emplaced as sills, dykes and igneous centres in the upper crust (Jerram & Bryan 2015). In cases where the volume of magma emplaced in sedimentary basins is high, these basins are commonly referred to as 'volcanic basins' (Fig. 10) (e.g. Jerram 2015; Abdelmalak et al. 2016;Jerram et al. 2016a). Volcanic basins are present along rifted continental margins and on lithospheric cratons (e.g. Coffin & Eldholm 1994), and represent vast basin areas that contain significant volumes of LIP-related igneous rocks and, in some cases, particularly where underplating has occurred and sill complexes exist, the intrusive component can be larger than the extrusive counterparts (basalt/lava) and pyroclastic deposits. Sill emplacement is characterized by the development of hydrothermal vent complexes that cut up through the basin and erupt at the surface through pipe structures (e.g. Fig. 10) (Svensen et al. 2006;Jerram et al. 2016a) that can be found both before and within the volcanic pile and which can indicate the relative timing of the sill emplacement (e.g. Angkasa et al. 2017).
Sill thicknesses can exceed 350 m, sandwiched between sedimentary sequences and associated contact metamorphic aureoles. Sills can also occur within the lava piles themselves (e.g. Hansen et al. 2011), but their identification is complicated within a similar host rock and also emphasizes a potential for the underestimate of the volume of the intrusive component. A selection of field photographs from several of the LIPs described in this work is shown in Figure 7. Contact metamorphism of the host sedimentary rocks cause devolatilization and the generation of H 2 O, CO 2 , CH 4 and SO 2 depending on the composition of the country rocks, including their content of organic carbon (e.g. Svensen et al. 2009). As a rule of thumb, the volume of heated sedimentary rocks is twice the sill volume, as suggested both by modelling studies and by mineral and maceral temperature proxies (e.g. Aarnes et al. 2011). These volatiles may reach the atmosphere via direct degassing in breccia pipes and hydrothermal vent complexes or by seepage through fractures and sediment permeability. In addition to the sediment degassing, basalt lava degassing releases a range of volatiles to the atmosphere, including CO 2 and SO 2 (e.g. Self et al. 2006).
A classic epitome of the LIP mass-extinction paradigm is the relationship between CAMP and end-Triassic mass extinction. Combined geochronological and stratigraphic data have evidenced that the Fig. 9. Subducted lithospheric slabs may interact with the margins of TUZO and JASON, triggering the formation of plumes. The cartoon is a profile approximately through the reconstructed Paraná-Etendeka LIP, North Madagascar-Seychelles and India (cf. Fig. 3) at around 135 Ma. The triggering slabs for plumes sourcing the Paraná-Etendeka LIP and contemporaneous kimberlites in South Africa and Namibia must clearly have been linked to subduction along the western margin of South America. The South Atlantic opened shortly after the Paraná-Etendeka LIP. two events coincide, but few constraints exist on how the CAMP played its forcing on the end-Triassic global environment. On the stratigraphic record, from worldwide localities, the end-Triassic biotic crisis is also associated to a negative carbon isotope excursion that reflects a disruption of the carbon cycle, for which the CAMP is considered the cause. Modelling shows that the sharp negative δ 13 C excursion associated with the extinction can be obtained by rapid release of very isotopically light carbon (e.g. Bachan & Payne 2016). Since carbon released by flood basalt magmatism is isotopically too heavy to drive such negative excursions, the source of isotopically light carbon must be sought in the degassing from heated sediments. In this view, the study of subvolcanic intrusions and the estimation of sill volumes are fundamental because the degassing from organic-rich sediments in contact with such magmatic structures is, to the best of our knowledge, the most efficient way to transfer volatiles from the lithosphere to the Earth's atmosphere during LIP events. The volume estimation of CAMP magmas presented here further highlights that an important portion of the CAMP is constituted by intrusive and subvolcanic products, and that the majority of them are hosted by the organic-rich sedimentary basins in Brazil.
There are still unresolved aspects related to the age of sills and dykes, their volumes, and emplacement rate, even though recent work has made advances in several LIPs (e.g. Jourdan et al. 2005Jourdan et al. , 2014Svensen et al. 2012;Blackburn et al. 2013;Burgess et al. 2015). These issues are challenging to resolve as sills and dykes are non-uniformly distributed in volcanic basins, and basin-scale 3D seismic data and boreholes are not always available. Still, we stress that current knowledge and data suggest that the subvolcanic intrusions in sedimentary basins hold a key to understand past environmental changes.

Summary
• In Palaeozoic and Early Mesozoic times, Gondwana LIPs and the majority of kimberlites were sourced by deep plumes from the margins of the African LLSVP (TUZO), one of two major stable thermochemical piles at the core-mantle boundary. Gondwana essentially existed from Cambrian to Jurassic times and in its lifetime three continental LIPs (Kalkarindji, CAMP and Karoo-Ferrar) directly affected Gondwanan continental crust at around 510, 201 and 183 Ma. • Two of these LIPs (CAMP and Karoo-Ferrar) assisted the break-up of Gondwana and Pangea, as witnessed by the opening of the Central Atlantic Ocean (splitting Pangea) and the West Somali and Mozambique basins (separating West and East Gondwana) during the Jurassic. • Soon after, the opening of the South Atlantic in the early Cretaceous from around 130 Ma took place only about 4-5 myr from the onset of the Paraná-Etendeka LIP, which affected vast areas in Brazil and parts of Namibia. • The architecture of the major Gondwana LIPs consists of a subvolcanic and a volcanic part. Subvolcanic sills and dykes represent a significant part of the LIP igneous volume, commonly emplaced in sedimentary basins (i.e. volcanic basins). Sills are commonly present throughout the basin stratigraphies and the sill volumes are thus scaled to the basin size. • A new assessment of the CAMP LIP shows that the sills have a volume of about 700 000 km 3 (based on an average volume estimate - Table  3), more than half of it present in the Brazilian basins. This suggests that sill emplacement and the related contact metamorphism and devolatilization probably contributed to the climatic change and mass extinction at the end of the Triassic.