Early Jurassic North Atlantic sea‐surface temperatures from TEX86 palaeothermometry

Early Jurassic marine palaeotemperatures have been typically quantified by oxygen‐isotope palaeothermometry of benthic and nektonic carbonate and phosphatic macrofossils. However, records of Early Jurassic sea‐surface temperatures that can be directly compared with general circulation model simulations of past climates are currently unavailable. The TEX86 sea‐surface temperature proxy is based upon the relative abundance of glycerol dialkyl glycerol tetraethers preserved in organic‐carbon‐bearing sediments. This proxy has been used extensively on Cretaceous and Cenozoic materials and, in one study, on Middle and Upper Jurassic sediments. Here, TEX86 is applied, for the first time, to Lower Jurassic (Sinemurian–Pliensbachian) sediments cored at Deep Sea Drilling Project Site 547 in the North Atlantic. The abundance of glycerol dialkyl glycerol tetraethers in these sediments is very low, despite biomarker and Rock‐Eval data suggesting that thermal maturity is, generally, low. Sea floor oxygenation and a high input of reworked terrestrially sourced organic matter may explain the low concentrations. For samples from which it was possible to quantify the relative abundance of glycerol dialkyl glycerol tetraethers, TEX86 values range from 0·78 to 0·88, equating to sea‐surface temperatures in excess of >28°C. These temperatures are broadly comparable with new general circulation model simulations of the Sinemurian and Pliensbachian stages and support the general view of a predominantly warm climate. The new proxy data suggest that, under favourable geological conditions, it is possible to extend the record of TEX86‐based sea‐surface temperatures back into the Early Jurassic.


INTRODUCTION
Although there is a consensus that the Mesozoic Era was characterized predominantly by warm 'greenhouse climates ' (e.g. Fischer, 1981;Frakes & Francis, 1988;Frakes et al., 1992), the spatial and temporal variability in climate during the Mesozoic is, generally, uncertain. The abundance of available records and the variety of proxies also tends to decrease with increasing age, such that there are relatively more data available for the Cretaceous than the Jurassic Period. These factors limit the understanding of past climate dynamics and sensitivity, particularly during intervals of major environmental change, and hinder the ability to compare the geological record with General Circulation Model (GCM) simulations of past climates. Many studies have compiled oxygen-isotope records from carbonate and phosphate fossils to reconstruct Mesozoic temperatures and temperature trends (e.g. Veizer et al., 1999;Jenkyns et al., 2002;Puc eat et al., 2003;Prokoph et al., 2008;Dera et al., 2011;Friedrich et al., 2012;Price et al., 2013;Korte et al., 2015). However, interpretation of these datasets can be complicated by a number of modifying factors including diagenesis, uncertainties regarding the oxygen-isotopic composition of local and global seawater, and the palaeoecology and physiology of the organism used (for example, the water depth inhabited by the organism, geographic and water depth migration, ontogeny and 'vital effects').
Reconstructing sea-surface temperatures (SSTs) for the geological past allows spatial and temporal palaeoclimatic variability to be investigated and directly compared with palaeoclimate simulations generated by General Circulation Models (GCMs). For the Cenozoic, numerous SST reconstructions have been generated, based on a wide range of proxies, including the geochemistry (d 18 O or Mg/Ca ratios) of planktonic foraminifera or the alkenone unsaturation index (U K 0 37 ). Although planktonic foraminifera become relatively abundant in the fossil record in the Middle Jurassic, at least in the European-Atlantic region (e.g. Caron & Homewood, 1983;Hart et al., 2002), their palaeoceanographic utility is limited to the mid-Cretaceous onwards when they become more abundant and with suitable preservation for geochemical studies (e.g. Leckie, 1989). Similarly, the oldest known alkenones are of Early Cretaceous age (Brassell & Dumitrescu, 2004), but their preservation is insufficient in Mesozoic sediments investigated to date to allow reconstruction of SSTs and the application of U K 0 37 is therefore limited to the last ca 55 Myr (Brassell, 2014). Hence, to reconstruct SSTs prior to the mid-Cretaceous, other proxies must be sought and tested.
Another approach for reconstructing SSTs is the TetraEther indeX of tetraethers consisting of 86 carbon atoms ('TEX 86 ') organic palaeothermometer (Schouten et al., 2002(Schouten et al., , 2003(Schouten et al., , 2013. Isoprenoid glycerol dialkyl glycerol tetraethers (isoGDGTs) are membrane lipids produced by Archaea with varying numbers of cyclopentane rings and, in the case of crenarchaeol and its regioisomer, a cyclohexane ring. Culture studies show that with increasing temperature the relative abundance of isoGDGTs containing higher numbers of cyclopentane rings increases, as does the relative abundance of the regioisomer of crenarchaeol (reviewed in Schouten et al., 2013); this is captured in the TEX 86 ratio (Schouten et al., 2002). Marine core-top studies suggest that TEX 86 best correlates with mean annual sea-surface temperature (e.g. Kim et al., 2010), although it is widely acknowledged that the depth interval of maximum isoGDGT production is not the sea surface itself and that the timing of peak production during a year is not consistent globally (reviewed in Schouten et al., 2013;Taylor et al., 2013): TEX 86 has been used to reconstruct SSTs through much of the Cenozoic and Cretaceous (e.g. Littler et al., 2011;Linnert et al., 2014) and, in one study, has been applied to immature Middle and Upper Jurassic sediments (Jenkyns et al., 2012). In order to extend the TEX 86 record to other Jurassic sites, it is necessary to identify organic-carbon-bearing sediments of low thermal maturity. Artificial maturation studies showed that the concentration of GDGTs within a sample and the TEX 86 ratio decrease rapidly with increased thermal maturation, while naturally thermally mature sedimentary rocks are devoid of GDGTs (Schouten et al., , 2013. In the study presented here, the occurrence of isoGDGTs is reported from Lower Jurassic (Sinemurian-Pliensbachian, ca 191 Ma) sediments using samples from Deep Sea Drilling Project (DSDP) Site 547 in the subtropical North Atlantic. Consideration of the thermal maturity of the organic matter and the distributions of isoGDGTs allows the first reconstruction of TEX 86 -based SSTs in sediments of this age, which are then compared with new GCM simulations for the Sinemurian and Pliensbachian. Importantly, the new data demonstrate the potential of TEX 86 for reconstructing SSTs in sediments much older than the Cretaceous, provided that depositional and diagenetic conditions were favourable for preservation and thermal maturity is low. Furthermore, the new model and proxy results provide palaeoclimate insights into an interval of the Jurassic that has not been simulated by climate models with modern, IPCC-class GCMs (cf. Chandler et al., 1992).

MATERIALS
Deep Sea Drilling Project (DSDP) Site 547 is located in the Atlantic Ocean, off the Moroccan coast ( Fig. 1A) and provides a discontinuous sedimentary record from the uppermost Triassic to Quaternary (Shipboard Scientific Party, 1984). During the Early Jurassic, the site was located on the margins of the proto-North Atlantic basin, which, during the Hettangian-Sinemurian (ca 201 to 191 Ma), was a large, low-latitude marine embayment connected only to the Tethys Ocean and north-west European shelf (e.g. Ziegler, 1990;Fig. 1B). The so-called 'Hispanic Corridor', connecting the Tethys Ocean with Panthalassa in the west, opened during the Sinemurian-Pliensbachian transition (ca 191 Ma), but only widened, with large-scale migration of faunas between the Tethyan and Pacific realms around the Pliensbachian-Toarcian transition (ca 183 Ma; Aberhan, 2001;Porter et al., 2013).
At Site 547, three wells were cored, of which only Hole 547B penetrated Lower Jurassic sediments (Fig. 2). The oldest sediments recovered in Hole 547B (below 932Á5 metres below sea floor; mbsf) were sandy mudstones containing gypsum veins with rare dolomite that were interpreted as being deposited in a non-marine environment and dated as Rhaetian-Hettangian (Shipboard Scientific Party, 1984). Between 932Á5 mbsf and 846Á0 mbsf in Hole 547B (most of lithological Subunit VIB; Shipboard Scientific Party, 1984), a diversity of sediment types were recovered and dated as Early Jurassic. The basal portion of this interval (932Á5 to 891Á0 mbsf) is a complex mixture of lithologies including mudstones, nodular limestones, limestones and dolomitic breccias. At the base (Hole 547B, cores 23 and 24) a laminated micrite resembling a stromatolite and breccias of stromatolite occur that could represent tidal-flat deposits (Winterer & Hinz, 1984). The overlying sediments were interpreted as being deposited in a pelagic, marine environment (radiolaria and calcareous nannofossils are present) with water depths of ca 100 m and influenced by down-slope transport (Winterer & Hinz, 1984). Between 891 and 846 mbsf, the sediments are interbedded claystones and nodular micritic limestones that were deposited in a pelagic environment, with less influence from downslope transport. Importantly, relatively immature, organic-carbon-rich sediments generally bearing 0Á5 to 2Á0% total organic carbon (TOC) have been reported (e.g. Shipboard Scientific Party, 1984). Samples were taken from cores 23 to 15 from Hole 547B with an emphasis on fine-grained, hemipelagic sediments. Lower Jurassic breccias and Middle-Upper Jurassic pelagic limestones recovered above 846 mbsf were not sampled due to their extremely low TOC contents (Shipboard Scientific Party, 1984) and likely unsuitability for organic geochemistry.

Nannofossil biostratigraphy
Calcareous nannofossils were analysed using simple smear slides and standard light microscope techniques (e.g. Bown & Cooper, 1998). Data were collected semi-quantitatively using a Zeiss photomicroscope (Carl Zeiss AG, Oberkochen, Germany) at 91000 magnification. Abundance and preservation categories are given in Table S1. Biostratigraphy is described with reference to the Jurassic NJ zones of Bown & Cooper (1998)   (PC: S1 + S2 + S3 CO + S3 CO 2 ) and the Residual Carbon (RC: S4 CO + S4 CO 2 ). Precision and accuracy of the Rock-Eval analyses was checked by regular measurement of the in-house standard SAB134 (a Lower Jurassic shale) and the international reference standard IFP160000. Repeated long-term analyses of the mineral carbon content of SAB134 yields a standard deviation of better than AE0Á37%. The long-term analyses of IFP160000 yield an average mineral carbon content of 3Á20%, with a standard deviation of 0Á02%, within error of the referenced value (3Á26 AE 0Á12%). For TOC, repeated analyses of SAB134 yields a standard deviation of better than AE0Á39%. The long-term average TOC value of IFP160000 is 3Á27%, with a standard deviation of 0Á04% (again within error of the reported value of 3Á28 AE 0Á14%). During the measurement of the samples from Site 547, 20 analyses of the in-house SAB134 standard indicate standard deviations of AE0Á06% for TOC, 0Á08% for mineral carbon-content, AE8 mgHC/gTOC for hydrogen indices (HI), AE2Á8 mgCO 2 /gTOC for oxygen indices (OI), and AE1Á7°C for T max .

Biomarkers
Approximately 4 to 7 g of powdered sample was ultrasonically extracted using methanol, then dichloromethane (DCM)/methanol (1:1, v/v) and, finally, 100% DCM. All extracts were combined and dried under N 2 at 40°C. Water was removed from the samples by passing the extracts [dissolved in DCM/methanol (3:1, v/v)] over a column containing anhydrous Na 2 SO 4 . Extracts were split into apolar and polar fractions by column chromatography, using hexane/DCM (9:1, v/v) and DCM/methanol (1:1, v/v) sequentially and Al 2 O 3 as the stationary phase. The polar extract was dissolved in hexane/propanol (99:1, v/v) and then filtered through a PTFE (polytetrafluoroethylene) 0Á45 lm filter. The polar extract was analysed at the University of Oxford using high-performance liquid chromatography/atmospheric pressure positive ion chemical ionization mass spectrometry (HPLC/APCI-MS), using an Agilent 1200 series LC, coupled to a G6130A single quadrupole mass spectrometer. The analytical protocol followed was as described in Schouten et al. (2007a). The abundance of isoprenoid GDGTs was measured in selective ion monitoring mode (m/z 1302, 1300, 1298, 1296, 1292, 1022, 1020, 1018, etc.). Ion peaks of the respective GDGTs were integrated to determine the relative abundance of each compound. The abundances of the isoGDGTs were used to determine the TEX 86 value using Eq. 1: where Cren 0 represents the regioisomer of crenarchaeol. Repeat analysis of an in-house standard following the same analytical sequences as samples reported here suggests an analytical precision on TEX 86 of AE0Á01 (1r, n = 10). The apolar fractions were re-dissolved in 10 to 50 lL hexane before analysis with a Thermo ISQ series single quadropole gas chromatographymass spectrometry (GC-MS) system (Thermo Fisher Scientific, Waltham, MA, USA) at the University of Bristol to determine the C 31 homohopane 17b,21b(H)/[17b,21b(H) + 17b,21a(H) + 17a,21b(H)] ratio (22R + 22S isomers combined), an indicator of the degree of thermal maturity (Mackenzie et al., 1980). After injection of 1 lL onto a Zebron-I non-polar column (50 m 9 0Á32 mm 9 0Á10 lm film thickness), the GC oven programme was: 70°C (1 min hold) to 130°C at 20°C min À1 , then to 300°C (held 24 min) at 4°C min À1 . The mass spectrometer continuously scanned between m/z 50 and 650. Hopanes were integrated using m/z 191.

Climate modelling
Global climate, including sea-surface temperatures, was simulated using the UK Met Office coupled ocean-atmosphere General Circulation Model HadCM3L version 4.5, configured with Sinemurian and Pliensbachian palaeogeographies. Lunt et al. (2016) gives experimental details for an equivalent set of Cretaceous, Palaeocene and Eocene simulations. The Jurassic simulations here are set up identically to those simulations, except for the use of Early Jurassic palaeogeographies and solar constant. In summary, the model has a resolution of 3Á75°in longitude by 2Á5°in latitude, in both the atmosphere and ocean, with 19 vertical levels in the atmosphere and 20 vertical levels in the ocean. The present study used Sinemurian and Pliensbachian palaeogeographies created by Getech Plc, using methods based on those of Markwick & Valdes (2004) and downscaled to the model resolution to produce the topographic and bathymetric boundary conditions required by the model. Atmospheric CO 2 in the model simulations has a value of 1120 ppm (four times pre-industrial values, 4 9 PAL), within the range of CO 2 estimates for the Early Jurassic based upon proxies and carbon-cycle models; as summarized in Breecker et al. (2010) and Franks et al. (2014). The insolation at the top of the atmosphere (Total Solar Irradiance, TSI) was calculated following Gough (1981), with a modern orbital configuration. Both simulations are run for a total of 1422 model years allowing the model to reach near equilibrium, and the climatic means presented here are calculated from the final 30 years of the simulations. For model-data comparisons, the modern latitude and longitude of Site 547 is rotated back to the Sinemurian and Pliensbachian based on the GETECH plate model, for consistency with the model simulations. Where the rotated site falls on land due to the course resolution of the model, the nearest ocean grid-box on the same latitude is utilized.

Nannofossil biostratigraphy
These results were originally presented in an unpublished PhD thesis (University College London) of Paul Bown (Bown, 1986) (Bown & Cooper, 1998). The data suggest that the studied sequence falls in the age range 194Á3 to 189Á0 Ma and is stratigraphically equivalent to the Oxynoticeras oxynotum, Echioceras raricostatum and Uptonia jamesoni ammonite zones, extending across the Sinemurian-Pliensbachian boundary (Gradstein et al., 2012).

Rock-Eval pyrolysis
Results of the Rock-Eval pyrolysis are shown in Figs 2 and 3, and (1r = AE0.4%, n = 49). There are no clear stratigraphic trends, with most of the variability resulting from changes in lithology between claystones and limestones. Inorganic carbon content varies from 1Á5 to 10Á6% with a mean of 5Á3% (1r = AE2.7%, n = 49). Due to the presence of both dolomite and calcite (Shipboard Scientific Party, 1984) inorganic carbon content is not converted to CaCO 3 . The T max values range from 370 to 538°C with an average of 428°C (1r = AE28°C, n = 49). Hydrogen indices (HI) range from 0 to 385 mg HC/gTOC. Oxygen indices (OI) range from 52 to 1125 mg CO 2 /gTOC.

Biomarkers
From the polar extracts, it was possible to quantify the abundance of isoGDGTs in nine samples (out of 68 attempted), all of which are Sinemurian in age. In many samples, GDGTs could not be detected or were present at extremely low abundance. Of the nine successful samples, three contained sufficient isoGDGTs to be analysed multiple times (Tables 1 and S3), yielding analytical reproducibilities for TEX 86 in these three samples of AE0Á01 (1r, n = 4), AE0Á02 (1r, n = 5) and AE0Á02 (1r, n = 4), comparable to the Sinemurian ~193 Myr. internal standard analysed in the same analytical sequences (1s = AE0Á01, n = 10). The TEX 86 values of the nine successful samples range from 0Á78 to 0Á88 ( Fig. 4 and Table 1) with an average of 0Á83 AE 0Á04 (1r, n = 9). It should be noted that at low concentrations there are likely to be analytical effects on TEX 86 values (Schouten et al., 2007a), which may lead to an additional source of error (probably less than AE0Á1). There is no clear stratigraphic trend in TEX 86 values but, interestingly, no samples below 894 mbsf yielded sufficient isoGDGTs for quantification of TEX 86 values. This depth coincides broadly with a lithological change at 891 mbsf (Fig. 4), from sediments containing abundant evidence for downslope transport (below 891 mbsf) to those interpreted as deposited in a hemipelagic environment with limited downslope contributions (above 891 mbsf; Winterer & Hinz, 1984). In every sample analysed, branched GDGTs (brGDGTs) were either below detection limit or insufficiently abundant to be quantified and hence Branched versus Isoprenoid Tetraether index (BIT index; Hopmans et al., 2004) values are not reported. From the apolar extracts, C31 homohopane bb/(bb + ba + ab) ratios were calculated for all of the samples that yielded isoGDGTS and for six additional samples that did not yield GDGTs. The C31 homohopane bb/(bb + ba + ab) ratios range from 0 to 0Á65, with an average of 0Á55. Of the 15 samples analysed, the majority (13) have C31 hopane bb/(bb + ba + ab) ratios higher than 0Á5 (Fig. 4). Figure 5 shows the modelled annual mean SST for the Sinemurian and Pliensbachian simulations. Globally, the highest SSTs in the Sinemurian and Pliensbachian are 32Á5°C and 33Á4°C, respectively, and occur in the Tethyan tropics (Fig. 5). On a regional scale the GCM simulates mean annual SSTs of 29Á02°C and 26Á97°C for the Sinemurian and Pliensbachian, respectively (quoted values are the mean of all ocean grid squares directly surrounding the nearest grid square to Site 547; Fig. 5). One difficulty in determining the modelled temperature for the location of Site 547 is that the palaeogeographic reconstructions have too coarse a geography to be able to represent the intricate archipelago and semi-restricted basin configuration depicted in Fig. 1. For both simulations, this complication may have a bearing on the estimated SSTs because they will be formed under local oceanographic conditions different from those that existed in reality. Sub-grid scale processes within the (semi-) restricted North Atlantic may have had a large dynamical control on local SSTs.

DISCUSSION
Maturity of the organic matter Schouten et al. (2004) demonstrated in artificial maturation experiments that increasing thermal maturity results in lower TEX 86 values (due to preferential loss of isoGDGTs containing two or more cyclopentane rings) and lower total abundances of isoGDGTs. With sufficient burial, thermal maturity increases to a level at which all isoGDGTs are lost. Thus, before interpreting the TEX 86 data from Site 547 it is important to consider the thermal maturity of the organic matter, especially given the age of the materials and the observation that many samples did not yield quantifiable amounts of isoGDGTs. The majority of samples analysed by Rock Eval (41 of 49) have T max values <435°C (Fig. 2), indicating that the organic matter is broadly immature. Three samples between ca 847 to 846 mbsf have exceptionally high T max values >470°C, suggesting that they are extremely mature. Because these samples occur at the shallowest burial depths of the sample set, it could suggest that reworked, more mature, organic matter is present in these samples. Many of the samples with high T max values are also characterized by high OI values (Fig. 3), indicating that the organic matter is very oxidized, which could also indicate an allochthonous origin. Unsurprisingly, these samples did not yield GDGTs. The GDGTs were only recovered from samples with T max values <430°C (Table 1). The C31 homohopane bb/(bb + ba + ab) ratios of >0Á5 in most GDGTbearing samples support the interpretation of low-maturity organic matter and suggest that the TEX 86 values reported here have not been compromised by catagenesis . The one sample analysed with a C31 hopane bb/(bb + ba + ab) ratio of 0 (i.e. containing no bb hopanes) did not yield GDGTs. Intriguingly, many samples did not yield sufficient GDGTs despite being immature as indicated by low T max values and, where available, C31 hopane bb/(bb + ba + ab) ratios >0Á5. More biomarker data are required to explore this relationship, but the preliminary findings suggest that maturity alone cannot explain the abundance of GDGTs in the Lower Jurassic sediments at Site 547. Neither does the presence or absence of GDGTs seem to be related to TOC; successful sediments span a range of TOC contents from 0Á5 to 1Á8%, and some high-TOC sediments did not contain detectable GDGTs. The low HIvalues and moderate OI-values of the GDGTbearing samples suggest that the organic matter is predominantly from a terrigenous source (Type III) or is partially oxidized, as previously suggested by Rullk€ otter et al. (1984). The very low abundances of branched GDGTs from Site 547 suggest that the contribution of soil organic matter to the terrigenous organic matter pool was very low, and that higher plant organic matter supply dominated.
The factors described here and other aspects of the depositional environment (for example, archaeal productivity, bottom-water oxygenation and sedimentation rates; e.g. Schouten et al., 2004;Littler et al., 2014) could have been more important than diagenesis or total rates of organic-matter accumulation in determining the concentration of GDGTs deposited during the Early Jurassic at Site 547. It is important to note that, although there is clear evidence for oxidation of organic matter, oxic degradation is not thought to significantly affect TEX 86 ratios (see discussion in Schouten et al., 2013). The coincidence between a decrease in the abundance of downslope transport of sediments at ca 891 mbsf and the increase in the abundance of GDGTs (as determined by the ability to calculate TEX 86 ratios) at ca 894 mbsf supports a primary oceanographic or depositional control on GDGT preservation and/or abundance.

Other possible factors influencing TEX 86 values
In addition to diagenesis, other factors, such as contamination by GDGTs sourced from soils (e.g. Weijers et al., 2006) or in situ production of GDGTs within the sediment pile by methanogenic and methanotrophic Archaea (e.g. Sinninghe Damst e et al., 2012;Zhang et al., 2011), can affect TEX 86 values and prohibit their interpretation as a proxy for SSTs. The absence, or extremely low abundances, of branched GDGTs in all samples suggests that the input of soilsourced GDGTs was extremely low (despite the possible input of terrestrially sourced organic matter: Rullk€ otter et al., 1984) and, therefore, was not an influencing factor on the TEX 86 values.
In situ production of GDGTs by methanogenic and methanotrophic archaea can be assessed through the use of the %GDGT-0 (e.g. Sinninghe Damst e et al., 2012) and the methane index (MI; e.g. Zhang et al., 2011). GDGT-0 (and to a lesser extent GDGT-1, GDGT-2 and GDGT-3) can be synthesized by sedimentary methanotropic Archaea (Koga et al., 1993;Weijers et al., 2006). Using data from culture studies, it has been suggested that if the %GDGT-0 (relative to %Cren) is greater than 67%, there could be an additional, nonwater column, source of GDGT-1, GDGT-2 and GDGT-3 (e.g. Sinninghe Damst e et al., 2012), thereby influencing the obtained TEX 86 value. The %GDGT-0 values for the Site 547 samples are all <15% (Table 1). The methane index (MI) relies upon the observation that normal marine sediments typically have MI values <0Á3, whereas sediments influenced by high rates of methane production have values >0Á5. The MI values for Site 547 are all <0Á2 (Table 1), suggesting marine conditions and a lack of GDGT production by methanotrophs.
Zhang et al. (2016) have suggested a new index for assessing the influence of non-thermal factors on TEX 86 that also identifies data that deviate from modern analogues. The Ring Index (RI) is a weighted average of the cyclopentane moieties and, in the modern core-top dataset, is significantly correlated with TEX 86 . Zhang et al. (2016) suggest that geological samples that deviate from this relationship may not be used for palaeothermometry with confidence, and use the offset between the RI value of a sample and the modern TEX 86 -RI relationship (termed 'DRI') to identify spurious samples, with absolute DRI (|DRI|) values >0Á3. The samples from Site 547 all have |DRI| values of ≤0Á3 and, therefore, do not seem in any way unusual compared to the modern dataset.

Early Jurassic sea-surface temperatures from DSDP Site 547
The absence of diagenetic or other confounding factors that may have compromised the GDGT distributions measured from Site 547, allows for the use of the TEX 86 data to reconstruct Sinemurian SSTs for the proto-North Atlantic. The TEX 86 values from Site 547 are higher than almost all values found anywhere in the modern ocean, with the exception of data from the Red Sea (Kim et al., 2010;Tierney & Tingley, 2015), suggesting that subtropical sea-surface temperatures were as warm, or warmer, than modern during the Early Jurassic. The low %GDGT-0 provides additional evidence for high SST (i.e. Schouten et al., 2002).
In the modern core-top calibration data, TEX 86 values from the Red Sea deviate slightly from the modern calibration and yield higher than observed SSTs, possibly due to the presence of a phylogenetically distinct archaeal population (Trommer et al., 2009). For geological samples, it is not yet easy to distinguish between ancient GDGT distributions that were the consequence of warmer than modern SSTs or ancient 'Red Sea-like' populations (e.g. Inglis et al., 2015;Zhang et al., 2015). In the case of many openocean records formed during the Cretaceous and Cenozoic it is unlikely that their high TEX 86 values are the consequence of restricted oceanographic conditions like those found in the modern Red Sea or Mediterranean, which might have led to atypical populations of Thaumarchaeota. However, the North Atlantic Basin in the Early Jurassic would, potentially, have had similarities with those modern basins, because it was situated at low latitudes and likely possessed only a single connection to the global ocean until the Early Pliensbachian (Aberhan, 2001), suggesting that the SSTs calculated from the GDGT data should be treated with some caution. Nonetheless, it is interesting to note that the nannofossil assemblages show relatively high diversities and abundances for this time interval (with the exception of barren samples), that are comparable with other open-marine Tethyan sections, suggesting normal conditions at Site 547 and good connectivity with the 'global' ocean.
A number of approaches exist for calculating SST from TEX 86 data (Kim et al., 2008(Kim et al., , 2010Schouten et al., 2013;Tierney & Tingley, 2014 and the different methodologies remain active areas of research and debate. Here, Early Jurassic SSTs are estimated using three approaches, all of which require extrapolation of modern SST-TEX 86 relationships. Firstly, a linear calibration (Eq. 2; Kim et al., 2008) is used that excludes modern core-top data from the Red Sea: Equation 2 is favoured over the linear calibration of Kim et al. (2010), because that calibration included cold-water data (<5°C) that may not conform to the linear SST-TEX 86 relationship observed at warmer temperatures. A linear relationship between TEX 86 and SST up to 40°C is supported by mesocosm experiments (Schouten et al., 2007b), yet the linear calibrations available (e.g. Kim et al., 2008Kim et al., , 2010) yield very high temperatures at high TEX 86 values, which some authors consider to be unrealistic (e.g. Hay & Floegel, 2012). Secondly, a Bayesian model approach is applied, BAYSPAR (Tierney & Tingley, 2014, that compares measured TEX 86 values with similar values in the modern core-top dataset to derive linear regression parameters: BAY-SPAR fully propagates uncertainties in the coretop data into resulting temperature predictions (Tierney & Tingley, 2014).
Thirdly, a logarithmic calibration (TEX 86 H , Eq. 3) from Kim et al. (2010) is applied, which yields lower estimates of SSTs at high TEX 86 values than the linear calibration. Although this approach is commonly used in many palaeoclimatic studies, it is not supported by mesocosm studies (reviewed in Schouten et al., 2013) and is associated with structural residuals at high values (Tierney & Tingley, 2014: Sinemurian SSTs are estimated (Table 1 and Fig. 4) to be between ca 33°C and 39°C using the linear calibration (Eq. 2); ca 28°to 32°C using BAYSPAR; and ca 31 to 35°C using the logarithmic calibration (TEX 86 H , Eq. 3). All three calibrations are within error of each other for each sample (Table 1). During the Sinemurian, Site 547 was at a palaeolatitude of ca 25°N and it was at ca 30°N in the Pliensbachian. Cretaceous and early Cenozoic estimates of SSTs from similar palaeolatitudes, based upon both TEX 86 and d 18 O of well-preserved planktonic foraminfera, are broadly in the same range (e.g. Forster et al., 2007;Bornemann et al., 2008;Littler et al., 2011;Linnert et al., 2014;Inglis et al., 2015), suggesting that the Early Jurassic SSTs are consistent with those derived for other past greenhouse worlds. North-west European Sinemurian-early Pliensbachian palaeotemperatures of ca 15 to 25°C have previously been reconstructed using d 18 O (e.g. Hesselbo et al., 2000;Dera et al., 2009;Suan et al., 2010;Gomez et al., 2016). Given that much of these data come from benthic or nektobenthic organisms (for example, bivalves, belemnites, brachiopods and demersal fish) that record sub-sea-surface environments, and are from higher palaeolatitudes (up to 40°N) than Site 547, a temperature offset between these different datasets and the TEX 86 data presented here is unsurprising. On the former point, for Cretaceous sediments where it has been possible to measure belemnite d 18 O and TEX 86 from the same stratigraphic sections deposited in hemipelagic settings, significant offsets of 4 to 14°C have been recorded (e.g. Mutterlose et al., 2010;Jenkyns et al., 2012).
The Early Jurassic SSTs predicted from the nearest grid squares (29°C and 27°C) in the climate model simulations of the Sinemurian and Pliensbachian (Fig. 5) are at the lower end of the range of average values predicted by TEX 86 (Table 1), The GCM temperatures are within the error margins predicted by BAYSPAR (Table 1). Irrespective of the choice of calibration, the similarity between the TEX 86 and GCM SSTs suggests that the GCMs provide a favourable approximation of Early Jurassic climates using a pCO 2 level of 1120 ppm (4 9 PAL), albeit with slightly lower average SSTs. The differences between the average SSTs calculated by TEX 86 from Site 547 and the GCM simulations could be due to a number of factors. The complex palaeogeography of the North Atlantic in the Early Jurassic may not be sufficiently well represented in the climate models and may not accurately reproduce ocean circulation on a local scale. Although 1120 ppm is within the range of pCO 2 estimates for the Early Jurassic based upon proxies and carbon-cycle models (summarized in Breecker et al., 2010;Franks et al., 2014), there is potential that CO 2 levels were slightly higher (or lower) than this, which would lead to higher (or lower) SSTs. If additional SST data could be generated for multiple Sinemurian and Pliensbachian localities and palaeolatitudes it would then be possible to determine whether the GCM simulations can reliably predict the correct patterns of SST differences between sites and at what pCO 2 level the GCMs predict the most similar temperatures to those estimated from the geological record.

CONCLUSIONS
The glycerol dialkyl glycerol tetraethers (GDGT) data from Site 547 indicate that, under the right depositional and burial conditions, it is feasible to extract and determine the distributions of these lipids for intervals of geological time much older than previously reported. However, the results here indicate that Site 547 is not an ideal site to reconstruct Early Jurassic sea-surface temperatures (SSTs) by TEX 86 because there is evidence to suggest that the action of syn-depositional processes, such as bottomwater oxygenation, may have reduced the abundance of GDGTs. Furthermore, it is possible that the Sinemurian North Atlantic was a warm, semi-restricted basin. Because some modern analogues of such basins are characterized by GDGT distributions that deviate slightly from the global TEX 86 -sea-surface temperature relationships, further work is required to ascertain whether the open-ocean relationships are applicable in this palaeogeographic setting. Nonetheless, the TEX 86 data indicate that mean sea-surface temperatures in the Early Jurassic were at least as warm as anywhere in the modern ocean and, probably, in excess of 28°C, comparable with similar palaeolatitudes during other intervals of the Cretaceous and Early Cenozoic. The reconstructed temperatures are generally slightly warmer than those indicated by climate-model simulations using an atmospheric CO 2 level four times greater than preindustrial values, but there are considerable uncertainties regarding Early Jurassic atmospheric pCO 2 levels. Modelling at higher CO 2 concentrations is likely to yield a better fit with the proxy evidence from Site 547.