Oceanic anoxic event 2 (OAE-2) occurring during the Cenomanian/Turonian (C/T) transition is evident from a globally recognized positive stable carbon isotopic excursion and is thought to represent one of the most extreme carbon cycle perturbations of the last 100 Myr. However, the impact of this major perturbation on and interaction with global climate remains unclear. Here we report new high-resolution records of sea surface temperature (SST) based on TEX86 and δ18O of excellently preserved planktic foraminifera and stable organic carbon isotopes across the C/T transition from black shales located offshore Suriname/French Guiana (Demerara Rise, Ocean Drilling Program Leg 207 Site 1260) and offshore Senegal (Cape Verde Basin, Deep Sea Drilling Project Leg 41 Site 367). At Site 1260, where both SST proxy records can be determined, a good match between conservative SST estimates from TEX86 and δ18O is observed. We find that late Cenomanian SSTs in the equatorial Atlantic Ocean (≥33°C) were substantially warmer than today (∼27°–29°C) and that the onset of OAE-2 coincided with a rapid shift to an even warmer (∼35°–36°C) regime. Within the early stages of the OAE a marked (∼4°C) cooling to temperatures lower than pre-OAE conditions is observed. However, well before the termination of OAE-2 the warm regime was reestablished and persisted into the Turonian. Our findings corroborate the view that the C/T transition represents the onset of the interval of peak Cretaceous warmth. More importantly, they are consistent with the hypotheses that mid-Cretaceous warmth can be attributed to high levels of atmospheric carbon dioxide (CO2) and that major OAEs were capable of triggering global cooling through the negative feedback effect of organic carbon-burial-led CO2 sequestration. Evidently, however, the factors that gave rise to the observed shift to a warmer climate regime at the onset of OAE-2 were sufficiently powerful that they were only briefly counterbalanced by the high rates of carbon burial attained during even the most extreme interval of organic carbon burial in the last 100 Myr.
 Here we tackle these problems by generating the first high-resolution tropical SST records for the CTBE from two different sites located in the proto-Atlantic. Ocean Drilling Program (ODP) Leg 207 recently recovered 30 to 95 m thick intervals of Cenomanian to Santonian OC-rich laminated mudstones (black shales) from Demerara Rise (western equatorial Atlantic, Figure 1) [Erbacher et al., 2004, 2005] and we investigated 20 m of carbonaceous C/T black shales at Site 1260. Additionally, a 13 m thick section of latest Cenomanian black shales almost devoid of carbonate was studied at Deep Sea Drilling Project (DSDP) Leg 41 Site 367 (Cape Verde Basin, offshore Senegal, eastern equatorial Atlantic, Figure 1), partly as a continuation of earlier work [Kuypers et al., 1999, 2002; Schouten et al., 2003]. The major objective of this study was to employ the high-resolution tropical SST records across the C/T transition generated here to test current hypotheses concerning paleogreenhouse climates and carbon cycling.
2. Material and Methods
 Both cored CTBE sections were investigated for bulk sediment total organic carbon (TOC) and carbonate (CaCO3) contents and δ13Corg, so that detailed stable organic carbon isotope stratigraphy could be used to correlate between sites. At ODP Site 1260 we applied two independent paleotemperature techniques. The first of these techniques is based on the application of oxygen isotope paleothermometry to foraminiferal calcite from clay-rich sediments that display a distinctive ‘glassy’ taphonomy (Figure 2) indicative of excellent preservation [Wilson et al., 2002]. However, in common with DSDP Site 367, poor carbonate preservation precludes the application of oxygen isotope paleothermometry from the core CTBE interval. Hence we have applied a second paleotemperature technique here and in stratigraphically adjacent intervals, i.e., an organic geochemical SST proxy, the tetraether index of 86 carbon atoms (TEX86) [Schouten et al., 2002].
 The TEX86 is based on a close correspondence between the ambient SST and variations in the composition of cell membranes of marine Crenarchaeota. A linear relationship between SSTs and distributions of crenarchaeotal membrane lipids (glycerol dialkyl glycerol tetraethers: GDGTs) is observed in a wide variety of Recent marine settings and Holocene surface sediments [Schouten et al., 2002; Wuchter et al., 2004]. This correspondence is also evident from growth and temperature adaptation experiments of extant marine Crenarchaeota carried out under laboratory conditions, where crenarchaeotal growth was observed in water temperatures up to 40°C [Wuchter et al., 2004; S. Schouten et al., Towards the calibration of the TEX86 paleothermometer for tropical sea surface temperatures in ancient greenhouse worlds, submitted to OrganicGeochemistry, 2006]. Salinity [Powers et al., 2004; Wuchter et al., 2004] and diagenesis [Schouten et al., 2004] have been shown to have no influence on TEX86 results. Analysis of particulate organic matter and settling particles recovered in sediment traps from several oceanic provinces demonstrate that the TEX86 signal that is detected in surface sediments is derived from the upper 100 m of the water column [Wuchter et al., 2005]. The stable carbon isotopic composition of crenarchaeotal lipids from OC-rich sediments of the Aptian/Albian OAE-1b [Kuypers et al., 2001] suggests that mid-Cretaceous Crenarchaeota used biochemical pathways similar to those employed by their extant counterparts. Furthermore, the GDGT compounds themselves are identical and their distributions in Cretaceous and present-day sediments are similar [Schouten et al., 2003; Jenkyns et al., 2004]. These lines of evidence imply that the proxy is applicable in ancient sediments under the precondition that the general correspondence between the TEX86 and SSTs was also valid in the geological past.
2.2. Material and Sites Studied
 ODP Leg 207 Site 1260 is situated on the gently inclined (∼1°) northwestern tip of Demerara Rise (9°15.95′N; 54°32.63′W, water depth 2549 m) [Erbacher et al., 2004], a NW–SE oriented prominent submarine plateau located offshore Suriname and French Guiana in the western equatorial Atlantic (Figure 1). Through the Albian to Campanian Demerara Rise was located in the core of the tropics [Erbacher et al., 2004; Suganuma and Ogg, 2006; see also 3.2.1] and the depositional water depth increased from shelf to upper bathyal [Erbacher et al., 2005]. The studied late Cenomanian to early Turonian succession (437–417 meters composite depth (mcd) shipboard splice [Erbacher et al., 2004]; see Tables S1 and S2 in auxiliary material) is dominated by dark, carbonaceous, laminated calcareous mudstones to marlstones (“black shales,” Figure 3) interbedded with light-colored, layered intervals composed of silt- to sand-sized carbonate particles (“limestones”).
 DSDP Leg 41 Site 367 is located in the Cape Verde Basin (Figure 1), offshore Senegal in the eastern equatorial Atlantic (12°29.2′N; 20°02.8′W, water depth 4748 m; estimated depositional water depth ∼3700 m and paleolatitude ∼5°N) [Lancelot et al., 1977; Kuypers et al., 2002]. Owing to deposition below the carbonate compensation depth [Lancelot et al., 1977], the investigated late Cenomanian section (637–650 meters below seafloor (mbsf), see Figure 4 and Table S3 in auxiliary material) comprises almost carbonate-free, laminated black shales that consist mainly of clay minerals mixed with terrigenous silicates and abundant OM. The hole was spot cored and coring gaps obscure the intervals between 645.8–643.45 and 636.55–625.5 mbsf. Bulk rock TOC and δ13Corg data included in this study from DSDP Site 367 (Table S3 in auxiliary material) partly stem from earlier investigations [Kuypers et al., 1999, 2002], which are augmented here by data generated from new samples to increase the sampling resolution. For reasons of consistency, completely new SST data based on the TEX86 were generated in this study from both available sample sets because the analytical method has been refined and improved since the previously published data from Site 367 [Schouten et al., 2003] were produced.
 At both investigated sites, ODP Site 1260 and DSDP Site 367, the Cenomanian/Turonian sediments generally are composed of large quantities of OM (up to ∼20 to 45% TOC, respectively) that is thermally immature and predominately of marine origin according to Rock Eval pyrolysis results [Herbin et al., 1986; Erbacher et al., 2004]. On the basis of biomarker analysis, terrestrially derived OM is present, but only as a minor constituent in the spectrum of the overall excellently preserved organic compounds [Kuypers et al., 1999; Forster et al., 2004]. The high sulphur content in the OM at both locations reveals an excellent preservational setting but also calls for an extraordinary, euxinic depositional paleoenvironment [Kuypers et al., 2002; Forster et al., 2004].
2.3. Geochemical Bulk Rock Parameters
 Stable carbon isotopes of organic matter (δ13Corg), total organic carbon (TOC) and carbonate (CaCO3) contents were determined at the Royal NIOZ (sample set stated in Table S2 of auxiliary material) by decalcifying weighed aliquots of powdered rock samples with 2N hydrochloric acid (HCl). The percentage of CaCO3 was determined by the dry weight loss after decalcification, assuming that all carbonate occurring in the samples is calcite. The decalcified sediments were analyzed in duplicate on a Carlo Erba 1112 Flash Elemental Analyzer coupled to a Thermofinnigan Delta Plus isotope mass spectrometer. Analytical errors range from 0.02 to 1% for TOC and 0.01 to 0.33‰ for δ13Corg (vs. VPDB). TOC on a carbonate-free basis (TOCcf) was calculated by the equation:
 At NOCS, a split of the powdered sediment samples was acidified by adding HCl in two steps, first a weak solution (∼5%), second a stronger solution (∼35%) and subsequently neutralized by dilution with deionized water (Milli-Q). Carbon elemental compositions of the acidified samples (AC) and the nonacidified samples (TC) were analyzed utilizing a Carlo Erba EA 1108 Elemental Analyzer (internal analytical precision ± 0.3%). TOC and the CaCO3 percentages were calculated using the following equations:
 The acidified samples were then analyzed for δ13C using a Euro Vector Elemental Analyzer using flash combustion coupled to a GV Instruments IsoPrime continuous flow mass spectrometer. Analytical precision is better than 0.15‰ VPDB.
2.4. The δ18O Analysis of Planktic Foraminiferal Calcite
 The δ18O and δ13C was analyzed on monospecies separates of Hedbergella delrioensis (∼25 individuals; 125–250 μm size fraction) and Heterohelix moremani (∼100 individuals; 63–125 μm size fraction), which were obtained from 1.5 cm thick half-round core slices taken at 10 cm intervals. Only exceptionally well preserved (“glassy,” Figure 2) foraminiferan specimens free of secondary calcite chamber infillings were analyzed. To minimize ontogenetic effects, only specimens of comparable size with adult chambers were analyzed. The samples were reacted under vacuum with 100% orthophosphoric acid at 90°C using a Micromass Multicarb Sample Preparation System coupled to a VG PRISM mass spectrometer. Results are reported versus VPDB, calibrated through the NBS-19 standard. The analytical precision for δ18O is better than ± 0.1‰ (corresponding to <0.5°C), and for δ13C it is ± 0.06‰ VPDB.
2.5. TEX86 Analysis
 For ODP Site 1260, subsamples (3 to 5 g dry mass) of the freeze-dried and powdered sediment samples (∼1.5 cm thick core slices taken every 10 cm and 2–5 cm in selected intervals) were extracted with dichloromethane (DCM)/methanol (9:1, v/v) using a Dionex Accelerated Solvent Extractor. Total lipid extracts were separated by Al2O3 column chromatography into apolar, GDGT and polar fractions by subsequent elution with hexane/DCM (9:1, v/v), DCM/methanol (95:5, v/v) and DCM/methanol (1:1, v/v). For DSDP Site 367, desulphurized (using Raney Ni) polar fractions [Kuypers et al., 2002] were chromatographically fractionated to yield a GDGT fraction as described above.
 GDGT fractions were analyzed for TEX86 [Schouten et al., 2002] by high-performance liquid chromatography/atmospheric pressure positive ion chemical ionization mass spectrometry (HPLC/APCI-MS) using an Agilent 1100 LC/MSD under conditions described previously [Jenkyns et al., 2004]. TEX86 values were determined using the GDGT compounds and the formula given in Schouten et al. . The analyses were carried out at least in duplicate, in critical intervals at least in triplicate. Reproducibilities of SSTs determined for this sample set were <0.22°C (Site 1260) and <0.33°C (Site 367).
3. Results and Discussion
3.1. Stratigraphic Framework
3.1.1. ODP Site 1260
 The CTBE (426.41–424.85 mcd) is evident from a positive δ13Corg excursion with a 6.6‰ amplitude (Figure 3c) and an increase in the TOC content by up to 14% (Figure 3a). Because age-indicative fossils are largely absent within this interval at Demerara Rise [Erbacher et al., 2004, 2005], the stratigraphic range of the CTBE is defined at ODP Site 1260 chemostratigraphically, namely by the δ13C excursion (shaded area in Figures 3 and 4). Following earlier concepts [Pratt and Threlkeld, 1984; Gale et al., 1993; Paul et al., 1999], we have subdivided the excursion into three main phases in the same way as did Kuypers et al.  (see also Table S2). Phase A represents the interval from the onset of the positive excursion to a first isotopic maximum, documented at Site 1260 by an increase from preexcursion δ13Corg values clustering around −28.3‰ to −22.4‰ between 426.41 and 426.21 mcd. Phase A would be equivalent to the “first buildup” phase of the isotopic excursion according to the division currently in use at the proposed European reference C/T section in SE England (Eastbourne, Sussex) [e.g., Paul et al., 1999; Gale et al., 2005; Voigt et al., 2006]. Phase B encompasses the ensuing modest decline followed by a second increase in δ13Corg values (426.21–425.66 mcd), resembling the “trough” and “second buildup” phases in Eastbourne [Paul et al., 1999; Gale et al., 2005; Voigt et al., 2006]. The upper part of phase B is characterized by a broad interval of high δ13Corg values (up to −22.0‰, 425.47 mcd) that is equivalent to the “plateau” (sensu Paul et al.  and Gale et al. , and Voigt et al. ). Phase C corresponds to the gradual return of high δ13Corg values related to the “plateau” to lower values in the range of −27.2‰ (above 424.85 mcd). Following earlier studies [Tsikos et al., 2004; Kolonic et al., 2005], we refer to OAE-2 here as equivalent to phases A plus B (latest Cenomanian) [Tsikos et al., 2004], while C represents the recovery phase above the C/T boundary (425.27 mcd). This chemostratigraphic framework is supported by calcareous nannofossil biostratigraphy at Site 1260, namely, the last occurrence of Axopodorhabdus albianus (late Cenomanian marker, 426.045 mcd) and the first occurrence (FO) of Eprolithus octopetalus (early Turonian marker, 424.91 mcd) [Hardas and Mutterlose, 2006]. Generally, this stratigraphic concept is consistent with other published CTBE sections [Tsikos et al., 2004; Erbacher et al., 2005], except that the FO of Quadrum gartneri is not a reliable biostratigraphic marker for the C/T boundary at Demerara Rise [Hardas and Mutterlose, 2006].
3.2. Conversion of Proxy Data Into Paleotemperatures
 At ODP Site 1260, δ18O values determined on planktonic foraminifera range from −4.9 to −3.5‰ (Table S1 in auxiliary material). TEX86 values range from 0.85 to 0.95 at ODP Site 1260 and from 0.84 to 0.95 at DSDP Site 367 (Tables S2 and S3, respectively).
where δw is the δ18O of the ambient Cretaceous seawater, a value that must be estimated. For all of our δ18O temperature calculations we started from the proposed value for the δw for mean seawater in a mid-Cretaceous world free of significant continental ice sheets [Shackleton and Kennett, 1975]:
 A global mean value for δw is appropriate for deep water but δw varies significantly in the modern oceans (∼−0.5 to +1.25‰) such that universal application of the mean value for δw would lead to substantial errors in temperature calculations. Therefore we adjusted the Cretaceous mean value to account for latitudinal changes in precipitation minus evaporation by using the equation of Zachos et al. :
where L is the local paleolatitude. Recent paleomagnetic studies indicate a progressive southward drift of Demerara Rise from 15° ± 5°N in the mid-Albian to 3.6° ± 2.6°N in the Campanian-Maastrichtian [Suganuma and Ogg, 2006]. However, the analysis and interpretation of projected paleolatitudes and trends is still the subject of ongoing work and the implications of the new results are partly not in full agreement with other published South American plate reconstructions. Thus, to be consistent with Norris et al.  and Wilson et al. , we estimated L here conservatively as 5°N. This leads to an estimated local δw for Demerara Rise of
that is equivalent to −0.53 (‰, VPDB) for CO2 equilibrated with this water [Bemis et al., 1998, Table 1]. This approach assumes that modern latitudinal gradients in δw apply to the mid-Cretaceous, an assumption that is reasonable for low-latitude sites (see discussion by Wilson et al. ).
 We consider the temperatures calculated using the above technique (Table S1) to be conservative estimates because we have not attempted to make additional adjustments to take account of the predicted lower seawater pH arising from the high Cretaceous pCO2 levels [Zeebe, 2001; Wilson et al., 2002]. This decision reflects the large uncertainties associated with estimating local paleolevels of aquatic pCO2, which would have been influenced not only by the global atmospheric value but also by levels of primary bioproductivity, stratification or upwelling conditions, etc. at Demerara Rise. Generally, absolute foraminiferal SSTs increase by about 2°–3°C if they are adjusted for elevated pCO2 at levels of respectively, 1000 and 2000 ppmv [Zeebe, 2001] (see Figure S1 and Table S1 in auxiliary material).
3.2.2. SST Estimates From TEX86
 Uncertainties in the absolute SST estimates from TEX86 analysis lie mainly in the calibration of TEX86 to SST. The current calibration of TEX86 for marine settings is based on Holocene core top sediments [Schouten et al., 2002; Wuchter et al., 2004] (see below) and runs from TEX86 values ranging from 0.3 to 0.7. Hence the values reported here require extrapolation of this calibration. We converted all TEX86 values generated in our study into SSTs following the approach of Schouten et al. . There, Cretaceous paleo-SSTs >28°C were estimated from TEX86 by extrapolation of data from Holocene marine surface sediments with annual SSTs > 20°C, resulting in the following equation:
The reason for using this calibration in contrast to that of the complete core top calibration is that there may be a deviation from a linear relationship at higher temperatures, although too few data points presently exist to confirm this trend. In theory therefore it is possible that the current TEX86 calibration for marine settings based on Holocene core top sediments deposited in SST ranges from 0° to 28°C is applicable for extrapolation to warmer SSTs. For reference, these SSTs are calculated by the current equation from Schouten et al.  updated by Wuchter et al. :
If equation (9) is employed to convert our TEX86 values, the resulting SSTs (Figure S1 and Tables S2 and S3) range from 37 to 44°C and are far above the present-day level (∼27°–29°C) and higher than based on oxygen paleothermometry (31°–37°C) even after pH adjustment (36°–40°C). On the contrary, if equation (8) is used, the resulting SSTs (32°–36°C) are well in the range of SSTs reconstructed here based on planktonic foraminifera, or even conservative relative to the foraminifera-based estimates that are adjusted for pH.
 Finally, we can also employ the correlation between TEX86 values determined on particulate organic matter (POM) derived from the upper 100 meters of the present-day marine water column and corresponding in situ temperatures [Wuchter et al., 2005].
If absolute SSTs are reconstructed using equation (10) (Tables S2 and S3), then SSTs range from 33° to 39°C. These temperatures exceed the range of non-pH-adjusted foraminiferal SST estimates but are still lower then those SSTs adjusted for pCO2 of 2000 ppmv (Figure S1 and Table S1).
 In most cases the absolute SST estimates agree well between the two proxies, the main exception being the general Holocene TEX86 calibration (Figure S1 and Table S2). TEX86 SSTs reconstructed based on this calibration line by far exceed all other SST estimates. From the spectrum of absolute temperatures that can be calculated from our TEX86 data set (Figure S1), we have chosen to present the most conservative estimates obtained by equation (8) in the text and Figures 3 and 4. However, it is important to note that our main concern is the pattern of relative change in paleotemperature versus time and this pattern and our interpretations of them are not contingent on the absolute temperatures in question.
3.3. Tropical Cenomanian to Turonian SST Records
3.3.1. ODP Site 1260
 Our conservative estimate of the paleo-SST seen at Site 1260 ranges from 31 to 37°C (Figure 3d and Tables S1 and S2; see section 3.2). Temperatures calculated from δ18O of “glassy” planktonic foraminifera show a greater overall range than those calculated from TEX86. The analytical precision of the two techniques is comparable (<0.5°C) so this greater range presumably reflects some combination of changes in surface water δ18O and/or carbonate chemistry (e.g., pH and CO3 ion [Wilson et al., 2002]) and greater variability in depth habitat within the water column integrated by the foraminifera [e.g., Fisher and Arthur, 2002]. With regard to depth habitat, no systematic interspecies offset in δ18O of Hedbergella delrioensis and Heterohelix moremani is evident in our data. Presumably, both generalistic species were living in well-oxygenated and mixed surface waters above a chemocline, which separated these waters from deeper, oxygen-depleted (temporally euxinic during OAE-2 [Kuypers et al., 2002]) waters. Foraminifera are conspicuously almost entirely absent between 428.47 to 423.59 mcd (Figure 3d). Likely explanations for this marked decline in foraminiferal abundances are extreme postdepositional dissolution within this interval, or environmental exclusion, e.g., through lack of tolerance to conditions of lowered seawater pH, salinity changes or a rise of the chemocline. Several observations at ODP Site 1260 support the former view but are less consistent with the environmental exclusion hypothesis. First, none of the few foraminiferal specimens that could be retrieved from this interval were suitable for isotopic analysis because of recrystallization and chamber infilling (with only one exception, Figure 3d). Secondly, changes in calcareous nannofossil assemblages are not evident in the late Cenomanian section prior to the CTBE [Hardas and Mutterlose, 2006]. Nevertheless, where stratigraphically overlapping data sets are available at Site 1260, conservative estimates for SST correspond well between the two paleotemperature techniques and this gives us added confidence in the absolute temperatures that we report here (Figure 3d).
 Several prominent features or marked changes occur within the SST record of Site 1260, and we label these time intervals T0–T5 (Figures 3 and 4 and Table S2). In Figure 3d, T0 marks the end of a gradual long-term temperature increase during the late Cenomanian (from ∼33°C to 35°C) and coincides with a stratigraphic level (428.47 mcd) where the quality of foraminiferal preservation starts to deteriorate markedly. Above follows a trough-shaped relative temperature minimum just below the CTBE where SST declines by about 2°C. The CTBE is associated with an abrupt shift to an even warmer thermal regime (up to approximately 35° to 36°C, Figure 3d). According to the chronostratigraphic framework of the global C/T reference section near Pueblo, Colorado, OAE-2 lasted for about 550 kyr [Keller and Pardo, 2004] (updated by Sageman et al. ). At Site 1260 this is equivalent to phases A and B (see 3.1.1: ∼150 and ∼400 kyr, respectively), implying that the earliest time interval (T1–T2) related to the warmer thermal regime represents ∼100 kyr. Significantly, the initiation of this rapid shift to a warmer thermal regime is coincident with the onset of OAE-2 as recorded in δ13Corg (base of CTBE phase A, Figure 3). Furthermore, the new warmer thermal regime is punctuated by an interval of markedly cooler temperatures. SSTs fall by up to 4°C from T2 to a minimum of ∼32°C at 426.06 mcd in ∼75 kyr and are quite variable, suggesting unstable paleoenvironmental conditions (see 3.4.2). This temperature minimum coincides at Site 1260 with an intermittent repopulation event by benthic foraminifera [Friedrich et al., 2006]. SSTs exceeding 35°C are seen again at 425.91 mcd. Thus we estimate a total duration for this late Cenomanian cooling event of about ∼150 kyr at Site 1260. The new warmer SST regime is reestablished (by T3, corresponding to about the middle of CTBE phase B, Figure 3), before the peak in δ13Corg is attained. SSTs remain remarkably high and stable during interval T3–T4 (35°–36°C; Figure 3d) while δ13Corg declines steadily through the post-OAE-2 recovery phase from heaviest values near the C/T boundary to a minimum at 424.63 mcd. A ∼2°C cooling during interval T4–T5 is accompanied by another brief benthic repopulation event [Friedrich et al., 2006] and preceded by a broad δ13Corg maximum (Figures 3c and 4b; see section 3.4.4). Above T5, well preserved foraminifera return and averaged early Turonian SSTs continuously decline from near 36°C toward 34°C (417.76 mcd).
3.3.2. DSDP Site 367
Figure 4 compares our Demerara Rise SST records with the new SST data generated from DSDP Site 367 (see section 2.2 and Table S3). Here the lower part of the CTBE was recovered from late Cenomanian carbonate-free black shales (643–637 mbsf, see section 3.1.2). The paleotemperature records from the two sites show remarkable similarities. At DSDP Site 367, the initial phase (A) of the CTBE, marked by increasing TOC and δ13Corg values, is also accompanied by a significant temperature rise (∼3°C, T1 to T2, Figure 4f) and this is followed by a pronounced stepwise cooling (by about 4°C). The A/B phase boundary of the CTBE coincides with an absolute temperature minimum at Site 367 (∼32°C, 640.11 mbsf). This is exactly the stratigraphic level where the completion of a major isotopic shift toward significantly heavier compound specific δ13C signatures of leaf-wax-derived long-chain n-alkanes has been observed [Kuypers et al., 1999]. This shift was interpreted as indicative of a floral change form C3- to C4-plant-dominated vegetation in NW Africa in response to a drastic reduction of atmospheric pCO2 [Kuypers et al., 1999]. The coincidence of both observations implies that pCO2 would have dropped to levels low enough to be favorable for C4 plant communities [Kuypers et al., 1999] while marine SSTs simultaneously declined to minimum values related to the cooling event (see section 3.4.2). Unfortunately, a coring gap above 636.55 mbsf obscures the younger part of the section so that the upper stratigraphic limit of the cooling event itself cannot be determined at this site, thus precluding any firm interpretations about the further evolution of pCO2. When compared to Site 1260, it appears more likely that the SST value of 35°C at 639.2 mbsf at Site 367 falls into the interval T2–T3 rather than corresponding to the base of T3–T4 (Figures 4c and 4f). The former interpretation would also be consistent with the observation that the n-alkane compound specific δ13C signatures recorded above 640.11 mbsf [Kuypers et al., 1999] did not decrease again, implying that pCO2 stayed still relatively low.
 Owing to stratigraphic incompleteness at Site 367, it is difficult to give absolute duration estimates for time intervals and events differentiated at this section.
3.4. Correlation and Implications of the Tropical SST Records
 Within the limits of the temporal resolution of our data, our records show that the onset of OAE-2 (T1) coincided with the initiation of a rapid shift from a warm Cenomanian to an even warmer Turonian SST regime across the equatorial proto–North Atlantic. Lower-resolution published δ18O records from the subtropical North Atlantic [Huber et al., 1999; Kuhnt et al., 2005], midlatitude shelf seas in NW Europe [Jenkyns et al., 1994; Voigt et al., 2004] and from the Southern Ocean [Clarke and Jenkyns, 1999; Huber et al., 1995] suggest that this shift also occurred outside of the equatorial Atlantic provided that these δ18O records reflect mainly a temperature signal and not salinity changes [Keller and Pardo., 2004] or bias by diagenetical overprint, e.g., related to facies changes [Paul et al., 1999]. In contrast to our tropical (Figure 4) and other, subtropical records [Huber et al., 1999; Kuhnt et al., 2005], no initial warming is indicated by the midlatitude OAE-2 record of Voigt et al. , where a pronounced shelf sea temperature increase (4°–5°C) occurred later, concomitant with the second increase or “buildup” phase in the δ13C excursion (corresponding to T3 and Figure 4). This may be due to the lower sample resolution in the latter study, especially when considering isotopic records from the English Chalk, revealing a trend toward lighter δ18O values at the base of the Plenus Marl Member [e.g., Jarvis et al., 1988; Jenkyns et al., 1994; Lamolda et al., 1994]. A pulse of hydrothermal activity associated with the emplacement of oceanic plateaus, particularly the Caribbean Large Igneous Province [Sinton and Duncan, 1997; Kerr, 1998; Snow et al., 2005] (E1 in Figure 4) is hypothesized to have synchronously fuelled this climatic warming (T1–T2, Figure 4) and marine bioproductivity by release of volcanogenic CO2 and nutrients [Brumsack, 2006].
3.4.2. Intermittent Cooling Event
 The ensuing marked cooling event observed in our records from the equatorial North Atlantic during the early stages of OAE-2 (T2–T3, Figure 4) requires an explanation beyond the local setting because it took place synchronously at two independent sites and in both cases, a pronounced cooling (by ∼4°C) to temperatures lower than pre-OAE-2 SSTs occurred. We note that the initial temperature decline started shortly before the first isotopic δ13C maximum was fully reached (uppermost part of phase A), while the main cooling trend continued during the upper part of this isotopic maximum and following trough phase (lower half of phase B, below T3 or isotopic “plateau”). This interval of lowered SSTs seen in our tropical records in the latest Cenomanian corresponds exactly to the chemostratigraphic position of a cooling episode during OAE-2 that has been widely recognized based on the pulse occurrence and southward migration of boreal faunal elements in midlatitude shelf seas in NW Europe [Jefferies, 1962; Jarvis et al., 1988; Jenkyns et al., 1994; Lamolda et al., 1994; Gale and Christensen, 1996; Voigt et al., 2004]. This observation has first been made in the Plenus Marl Member of the English Chalk and age-equivalent strata located in the Anglo-Paris Basin in SE England and France [Jefferies, 1962] and hence was termed the “Plenus Cold Event” [Gale and Christensen, 1996]. It seems unlikely that the close stratigraphic correspondence between the European cooling event and those observed contemporaneously in both our tropical SST records here can be explained by a coincidence.
 We interpret this late Cenomanian cooling phase as a direct consequence of the pCO2 decrease [Freeman and Hayes, 1992; Kuypers et al., 1999] caused by extensive black shale deposition during OAE-2 [Arthur et al., 1985, 1988; Jenkyns et al., 1994]. Yet, the range of latest Cenomanian paleotemperatures reconstructed here (33°–35°C) for Site 367 clearly exceeds those observed from the present-day equatorial SSTs (28°–29°C), implying that average tropical air temperatures should have been warmer than today in the same measure. In the light of the Kuypers et al.  hypothesis, our new warmer SST estimates suggest that the crossover from C3 to C4 plants (see section 3.3.2 and their Figure 3) could already have occurred at slightly higher CO2 concentrations (e.g., ∼600 ppmv at ∼35°C).
 It is quite remarkable that, in comparison to the remainder of the TEX86 record at Site 1260, SST values show an unusually high degree of temporal variability (∼32°–35°C) during this particular colder interval (T2–T3, Figures 3d and 4c). Between 425.6 and 425.9 mcd, a marked decline of TOC (given on a carbonate-free basis in Figure 4a) is related to a carbonate rich horizon that has no equivalent at DSDP Site 367 because of the general absence of carbonate at that site [Lancelot et al., 1977]. Just below this horizon (near 426.0 mcd), repopulation by benthic foraminifera was recognized [Friedrich et al., 2006]. This short-term repopulation event commonly occurs within the lower third of the OAE-2 interval at the Demerara Rise (ODP Sites 1258–1261) [Friedrich et al., 2006]. Because the unusual variability in the TEX86 record appears to be restricted to a stratigraphic interval marked by a temperature decline and simultaneous reoxidation of the bottom water zone as indicated by the presence of benthic foraminifera, it appears likely that these changes reflect recurrent conditions of instability in the paleoenvironment (i.e., water column stratification intermittently with enhanced mixing, see below). A similar picture of rapidly varying lighter and heavier δ18O values emerges from several earlier studies on the Plenus Marls [Jarvis et al., 1988; Jenkyns et al., 1994; Lamolda et al., 1994], although it cannot be ruled out that the large variability of δ18O values observed there is at least partially linked to cyclic lithological changes and related different degrees of diagenetic overprint [e.g., Jenkyns et al., 1994; Lamolda et al., 1994]. Unfortunately, this hypothesis cannot be further tested at DSDP Site 367 (Figure 4) because of stratigraphic incompleteness and the general absence of carbonate there.
 Simultaneously with the pronounced cooling phase observed at both investigated paleotropical sites, recolonization by benthic foraminifera occurred not only at the Demerara Rise [Friedrich et al., 2006] but also during the lowest Whiteinella archaeocretacea zone in North America (“benthonic zone,” ∼lower Sciponoceras gracile ammonite zone [Eicher, 1969; Eicher and Worstell, 1970] (see Leckie et al.  for a review)) and Morocco [Keller and Pardo, 2004], indicating geographically extensive bottom water reoxidation. If this process took place on large enough scales globally, support would be gained for the suggestion that the cooling episode caused sudden changes in the marine paleoenvironment and probably also affected oceanic circulation by introducing instability into the vertical water column structure. Enhanced water mass mixing, especially as recurring events, would not only transfer oxygen and heat from surface to deeper waters but also should in turn aid the reintroduction of nutrients and degradation products of organic matter (trapped in subsurface waters) into the photic zone. As a consequence, this “recycling” process could have refueled the primary bioproductivity in the surface waters, potentially setting off an oscillating productivity-anoxia feedback mechanism (sensu Ingall et al.  and Murphy et al. ) that would explain why OC sequestration via black shale deposition persisted while bottom water reoxidation did not. Further evidence beyond our present geochemical data set, e.g., for the occurrence of intermittent anoxia [e.g., Simons and Kenig, 2001; Kenig et al., 2004], but also evidence for lowered surface to bottom temperature gradients in high resolution would be needed to further constrain the above assumptions (e.g., by analogy to the late Albian OAE-1d [Wilson and Norris, 2001]). However, evidence for a general increase of benthic water temperatures and a lowered surface to bottom temperature gradient during OAE-2 is presently only available from two subtropical C/T sections in the North Atlantic (DSDP Site 551 [Gustafsson et al., 2003] and ODP Site 1050 [Huber et al., 1999]) where stratigraphic incompleteness precludes interpretations in the interval of interest or on a high-resolution level.
3.4.3. Establishment of a New Warm Climate Regime
 On the basis of chemostratigraphic correlation, the warming trend, which leads to SSTs that are again in the range of 35°–36°C from T3 onward, appears concurrent with a second, even stronger eruption pulse of the Caribbean oceanic plateau (E2 in Figure 4) [Snow et al., 2005]. The latter eruption pulse could have led to enhanced CO2 emission and thus even further enhanced global warming. A recent study [Kuroda et al., 2007] indicates that a period of enhanced volcanic activity might have occurred slightly earlier than postulated by Snow et al. , before the second δ13C maximum marking the base of the isotopic plateau phase. This could explain the lighter δ13Corg signatures and the warming trend observed in the uppermost part of interval T2–T3. Alternatively, lighter δ13Corg values in the lower part of CTBE phase B concomitant with the cooling event (Figures 4b and 4c) might have been partially linked to weakened ocean stratification and processes of bottom water reoxidation (see section 3.4.2). Once reestablished, the persistence of the new warmer climate regime during the final stage of the CTBE (C) and beyond might indicate decreasing global rates of marine OM burial and/or, lasting elevated CO2 flux to the atmosphere from enhanced global volcanism.
 Elsewhere, enhanced volcanic and hydrothermal activity on a global scale has been invoked by numerous studies as a trigger mechanism for OAE-2 and to explain oceanic anoxia and marine productivity increase [e.g., Schlanger et al., 1987; Larson 1991; Sinton and Duncan, 1997; Kerr, 1998; Snow et al., 2005; Brumsack, 2006]. Alternatively, the late Cenomanian opening of an Equatorial Atlantic Gateway (EAG), thus the full establishment of a deep water connection between the northern and southern proto–Atlantic Ocean, has been postulated as a trigger for the pronounced paleoceanographic and paleoenvironmental changes associated with the CTBE [e.g., Tucholke and Vogt, 1979; Summerhayes, 1981; Wagner and Pletsch, 1999; Poulsen et al., 2001; Kuypers et al., 2002]. Unfortunately, however, the uncertainties associated with our age estimates for mid-Cretaceous volcanic eruption events and gateway configuration changes are considerable, making it difficult to test these hypotheses. Several modeling studies [Poulsen et al., 2001, 2003] suggest that paleogeographic changes associated with the opening of the EAG deep water passage in the Atlantic would have had a pronounced impact on global oceanic circulation and initiated large regional changes in seawater temperatures and salinity in the northern and southern part of the Atlantic. Voigt et al.  have invoked the EAG opening hypothesis and related paleoceanographic changes [Poulsen et al., 2003] to explain the marked temperature rise that they observe after the “Plenus Cold Event” in midlatitudinal European CTBE sections and also the ensuing relative stable and warm shelf seawater temperatures in the early Turonian. Recent investigations on Upper Cretaceous benthic foraminiferal assemblages at Demerara Rise locality by Friedrich and Erbacher  argue for a progressive opening of the EAG leading to a full deep water passage between the North and South Atlantic Ocean from early Campanian onward. This is in line with the postulation of a modeling study on paleoupwelling in the Atlantic [Handoh et al., 2003] that such a deep water connection was not established until post-Cenomanian time. However, even if EAG opening played no role with respect to triggering paleoceanic and climatic changes during the CTBE, it still seems likely that major changes in the vertical oceanic water column structure took place during the CTBE in the tropical Atlantic, especially when considering the major, eustatic sea level rise that occurred through that interval [e.g., Schlanger and Jenkyns, 1976; Jenkyns, 1980; Arthur et al., 1987, 1988; Leckie et al., 2002; Keller and Pardo, 2004] (recent review by Voigt et al. ).
3.4.4. Post-OAE-2 Events
 It is noteworthy that another, ∼2°C cooling period is evident from the TEX86 SST record at Site 1260 well above the CTBE between T4–T5, albeit based on low sampling resolution (Figures 3 and 4 and Table S2 in auxiliary material). A second repopulation event of benthic foraminifera [Friedrich et al., 2006] is observed at this time. The interval is preceded by a broad δ13Corg maximum of about 3‰ magnitude that is mostly developed within a thick carbonate bed (424.3–423.9 mcd) and characterized by remarkably constant TEX86-based SSTs (close to 35°C). In this succession, these events have not been described from other C/T sections besides those at Demerara Rise, and thus their paleoceanographic implications are not known. However, a brief positive δ13C excursion (bulk carbonate) situated closely above the CTBE was recognized by Jarvis et al.  in the English Chalk and termed the “Holywell Isotopic Event”. In a quite similar stratigraphic position, a positive spike of ∼3‰ is present in the central Western Interior Seaway (Colorado, Portland Core δ13Corg record) [Sageman et al., 2006]. Thus, if these are synchronous features, it might be that these positive isotopic spikes could be related to a global rather than a local cause. Elevated surface water productivity during the deposition of the carbonate layer, may even be enhanced by water column stratification leading to an increased surface to bottom water isotopic gradient, and could serve as an explanation for the observed increase in δ13Corg values at Site 1260. However, the trigger for this simultaneous change in primary bioproductivity and lithology, while SSTs remained unaffected remains an issue of speculation. If the observed cooling during T4–T5 was linked to enhanced carbon sequestration in the preceding interval outlined by the δ13C maximum and not related to another, local cause, it should be traceable in other regions, something that remains to be investigated.
 High-resolution SST records from two different paleoequatorial Atlantic sites based on two independent paleotemperature proxies show that the C/T transition falls within the warmest time interval of the mid-Cretaceous and, in keeping with greenhouse theory, even our most conservative estimates of paleotemperature exceed the range of present-day tropical SSTs (by ∼3° to 9°C). Whatever the underlying factor(s) responsible, we observe an initiation of a new warmer climate regime in lockstep with the onset of OAE-2. Our records also show that a temporary marked cooling event occurs during OAE-2. OAE-associated cooling was likely geologically short-lived (≤150 kyr). On the basis of correlation to and evidence from C/T sections outside the tropics, we interpret this cooling phase as a global event as documented by paleoenvironmental changes due to widespread reoxidation of bottom waters, probably also associated with unstable paleoceanographic conditions. The subsequent rewarming and establishment of stable warm conditions that characterize the new warm climate regime and persist post-OAE-2 into the early Turonian, imply that even the high rates of OC deposition attained during one of the most extreme carbon cycle perturbations in Earth history were insufficient to counterbalance lasting elevated contemporaneous CO2 influx to the atmosphere.
 We thank the Deep Sea Drilling Project and Ocean Drilling Program (ODP) for providing samples; the ODP Leg 207 Shipboard Scientific Party; W. Hale, and A. Hetzel for help with sampling; J. Ossebaar (NIOZ) for analytical assistance, E. C. Hopmans and M. Kienhuis for support with HPLC analysis; P. Hardas, J. Mutterlose, H.-J. Brumsack, and A. Hetzel for sharing data and discussions; J. Erbacher, O. Friedrich, N. Ohkouchi, and S. Voigt for discussions; and M. M. M. Kuypers for support and sharing ideas. J. Erbacher, H. C. Jenkyns, and E. J. Rohling are thanked for their efforts in improving and editing this manuscript; three anonymous reviewers gave valuable input on an earlier manuscript version. This work was supported by the European Community's Improving Human Potential Program (HPRN-CT-1999-00055, C/T-NET) and the Research Council for Earth and Life Sciences (ALW 812.03.002, Netherlands).