It is predicted that by the year 2300, the atmospheric CO2 concentration will exceed ~2000 ppmv (Caldeira & Wickett, 2003; Mikolajewicz et al., 2007), corresponding to a release of 4000 x 1015 g carbon (PgC) by fossil fuel emissions and land use changes since the beginning of the industrial revolution. The anthropogenic carbon will eventually sequester on time scales of 100,000 yrs as organic carbon into the ocean and land biosphere and as CaCO3 into the geosphere (Archer et al., 1998). This carbon transfer in the atmosphere-ocean system is comparable to that at the Paleocene-Eocene boundary (55 Ma), when a massive release of carbon into the climate system led to a prominent global warming event referred to as the Paleocene-Eocene Thermal Maximum (PETM). The PETM is characterized by a major (>3.0‰) negative carbon isotope excursion, documented in marine and terrestrial fossils (e.g. Koch et al., 1992; Kelly et al., 1998; Handley et al., 2008), and a worldwide seafloor carbonate dissolution horizon (e.g. Bralower et al., 1997; Lu et al., 1998; Schmitz et al., 1996; E. Thomas et al., 2000) as well as shoaling of the lysocline and carbonate compensation depth (Zachos et al., 2005). These changes are consistent with the release of more than 2000 PgC of isotopically depleted carbon into the ocean-atmosphere system within less than 10,000 years (Panchuk et al., 2008; Zachos et al., 2007, 2008), pointing to a greenhouse gas-driven warming (see Fig. 1). Recent estimates from Cui et al. (2011) indicate a slow emission rate of 0.3-1.7 PgC yr-1 as compared to the present-day emission of carbon dioxide of ~9.9 PgC yr-1 from fossil fuel emissions (Boden et al., 2010) and land-use changes (Houghton, 2008). Surface temperatures increased by 5°C in the tropics (Tripati & Elderfield, 2005; Zachos et al., 2005) and mid-latitudes (Wing et al., 2005), and by 6-8°C in the ice-free Arctic and sub-Antarctic (Hollis et al., 2009; Kennett & Stott, 1991; Moran et al., 2006; Sluijs et al., 2006, 2007, 2008a, 2011; E. Thomas et al., 2000; Weijers et al., 2007), and deep-sea temperatures increased by 4-6°C (Tripati and Elderfield, 2005; Zachos et al., 2008), relative to Paleocene temperatures (see Fig. 1). At the same time, large-scale changes in the climate system occurred, for example in the patterns of atmospheric circulation, vapor transport, precipitation (Robert & Kennett, 1994; Pagani et al., 2006a; Brinkhuis et al., 2006; Sluijs et al., 2008a, 2011; Wing et al., 2005), intermediate and deep-sea circulation (Nunes & Norris 2006; D.J. Thomas, 2004; D.J. Thomas et al., 2008) and a rise in global sea level (Sluijs et al., 2008b; Handley et al., 2011). The sea level rise is caused by various factors, including thermal expansion, decrease in ocean basin volume, decrease in mountain glaciers, as well as local tectonic changes. Topography and bathymetry during the PETM differed significantly from today with respect to the distribution of landmasses, sizes of ocean basins and width and depth of seaways.
2. Climate change and variability at the beginning of the PETM
The causes leading to the warming event at the Paleocene-Eocene boundary are still controversial (Fig. 2).
One possible sequence of events inferred from paleoproxies begins with a volcanically induced greenhouse gas (water vapor, CO2, CH4, and other constituents) increase that would have produced a global increase in surface temperature (Bralower et al., 1997; Kennett & Stott, 1999; Sluijs et al., 2007, 2011; E. Thomas et al., 2000). Various climate-modeling studies have investigated the warming event at the Paleocene-Eocene boundary in response of the elevated greenhouse gas concentrations. These studies utilized atmospheric general circulation models (Sloan & Barron, 1992; Sloan & Rea, 1995; Huber & Sloan, 1999; Shellito et al., 2003; Shellito & Sloan, 2006), ocean general circulation models (Bice et al., 2000; Bice & Marotzke, 2002), or more recently coupled comprehensive climate models (Heinemann et al., 2009; Huber & Sloan, 2001; Huber & Caballero, 2003; Huber & Caballero, 2011; Lunt et al., 2010; Shellito et al., 2009; Winguth et al., 2010) to simulate the mean climate and its variability during the Eocene, but they have not been able to reproduce the high temperatures of the PETM in the high latitudes, and were controversial regarding the cause of this warming (Pagani et al., 2006b; Zeebe et al., 2009).
Some of the more recent studies have investigated the climate feedbacks with a sequence of different greenhouse gas concentrations (e.g. Heinemann et al., 2009; Lunt et al., 2010; Winguth et al., 2010). In the following, we summarize key findings of the paper of Winguth et al. (2010), using a complex earth system model, the comprehensive Community Climate System Model version 3 (CCSM-3; Collins et al., 2006), in order to investigate PETM climate feedbacks in response to rises in the greenhouse gas concentrations. Huber & Caballero (2011) used the same model, but with a different dust concentration in the atmosphere. The simulated increase by 2.5°C from 4xCO2 to 8xCO2 in CCSM-3 could be explained by CO2 emissions due to enhanced volcanic activity at the beginning of the PETM (Fig. 3).
Surface temperatures in the tropics rise by only ~2°C from 4xCO2 to 8xCO2, in agreement with temperature reconstructions (Pearson et al., 2007) and with future climate predictions (IPCC, 2007) of a more extreme warming at high latitudes vs. low latitudes in a warmer world. Temperature increase over land exceeds that over the ocean (Fig. 4) due to reduced latent heat fluxes and lower heat capacity. Over the continents, the 30C isotherm in the 8xCO2 simulation reaches up to 30 latitude, about 5 more poleward than for the present-day simulation. Maximum simulated temperatures, comparable to extreme temperatures in the present-day Sahara, are simulated over subtropical Africa and South America (~50°C for 8xCO2), resulting in warm sea-surface temperatures in the adjacent oceans through advection. Simulated minimum surface temperatures (for 8xCO2) are between 3°C and 7C over the Arctic and about -10C over northeast Asia.
While data-inferred paleotemperatures are relatively well represented in the tropical regions, a significant bias between model results and data remains for the Arctic Ocean (Sluijs et al., 2006) and for the area around New Zealand (Fig. 5, Waipara River; Hollis et al., 2009). The bias in the northern polar region (IODP core 302 A; Sluijs et al., 2006) is of complex nature and could for example be associated with the concentration of cloud condensation nuclei used in CCSM-3 (Huber & Caballero, 2011; Kump & Pollard, 2008), the uncertainties in paleolocations (N-S position, or distance form shore), or with skewing of data towards summer temperatures (Sluijs et al., 2006). The causes for model-data discrepancies at high southern latitudes remain controversial.
A positive climate-carbon cycle feedback loop leading to further PETM warming due to destabilization of methane hydrates is shown in Fig. 2. There is sufficient evidence from various sites around the globe, including the New Jersey shelf (Sluijs et al., 2007), the North Sea (Bujak & Brinkhuis, 1998; Sluijs et al., 2007), the Southern Ocean (Kennett and Stott, 1991), and New Zealand (Hollis et al., 2009) that 13C-depleted carbon in the form of isotopically light CO2 and/or CH4 was released from the sea floor (Dickens et al., 1995, 1997; Higgins & Schrag, 2006; Pagani et al., 2006a) or from wetlands (Pancost et al., 2007) into the atmosphere-ocean-biosphere system (Sluijs et al., 2007; Bowen and Zachos, 2010). As shown in Figs. 3 and 4, such a change in the radiative forcing from 8xCO2 to 16xCO2 leads to a simulated additional warming of ~2°C globally, with 4°C at the poles, 5C over South America and South Africa, and 2°C at the equator. For the Southern Ocean, cool water masses moderate the climate over the polar southern hemisphere, so that south of 60, the temperature increase in response to the increase of CO2-radiative forcing is smaller than in the northern hemisphere (Fig. 4b). In mid-latitudes, the bias between the 16xCO2 simulation and reconstructed PETM surface temperatures is reduced compared to simulations with a lower atmospheric CO2 level with high dust concentration; the values are comparable to the 8xCO2 scenario with lower dust concentration in Huber & Caballero (2011). For the tropics, evidence from fossil remains of a giant boid snake in northeastern Colombia (Head et al., 2009) and modeling studies (Winguth et al., 2010; Huber & Caballero, 2011) support warm average temperatures of 30-34C.
Rapid warming at the beginning of the Eocene has been inferred from the widespread distribution of dinoflagellate cysts (or dynocysts). The abundance of one dynocyst species, Apectodinium, dramatically increased at different locations worldwide (Bujak & Brinkhuis, 1998; Crouch et al., 2001; Heilmann-Clausen & Egger, 2000; Sluijs et al., 2007), implying a change in environmental conditions such as warmer sea surface temperatures and increased food availability in form of phytoplankton (Burkholder et al., 1992) due to increased nutrient delivery by weathering (Ravizza et al., 2001; Zachos & Dickens, 2000) and erosion (Fig. 2).
The climate-carbon cycle feedback associated with an increase in greenhouse gases (Fig. 2) might also have been enhanced by an increase in the atmospheric water vapor fluxes (Figs. 6 and 7); for instance, latent heat flux by evaporation and precipitation rises with the warming of the surface (Fig. 6). Higher precipitation and lower sea surface salinity values are derived for the Arctic from isotopic measurements as well as from dinocyst assemblages (Pagani et al., 2006a; Sluijs et al., 2008a). The enhanced precipitation at high latitudes is consistent to patterns simulated for future climate scenarios (e.g. Cubasch et al., 2001; Meehl et al., 2006; Mikolajewicz et al., 2007). For the southern high latitudes, a simulated increase in precipitation is confirmed by clay-mineral indicators from the Antarctic continent, pointing towards humid conditions at the PETM (Robert & Kennett, 1994). Compared to present-day, differences in the geography and mountain height cause remarkable changes. For instance, a higher than present-day ratio of tropical land-to-ocean area at the PETM reduces the tropical ocean surface and hence the oceanic source of atmospheric moisture (Barron et al., 1989). This change in tropical surface area not only reduces significantly tropical precipitation, but also poleward moisture transport from the tropics. However, increase in precipitation by higher than present-day greenhouse gases counteracted this effect during the PETM.
An initial increase in CO2 in the atmosphere by volcanic outgassing would have increased the strength of the hydrological cycle. Model simulations suggest that the subtropics at ~30° became drier and that precipitation at 60° increased significantly (Fig. 6), which is consistent to future climate projections (IPCC, 2007). Over North America during summer, the simulated total amount of rainfall decreases from lower to mid-latitudes in response to a northward-directed monsoonal moisture transport over the Mississippi watershed from the Gulf (Sewall & Sloan, 2006; Winguth et al., 2010). Sedimentary records from the mid-latitudes of the North American continent have produced conflicting evidence for hydrological changes in this region. For example, a ~25% increase in relative humidity for the northern continental mid-latitudes (Bighorn Basin, Wyoming, paleolatitude ~49 °N) has been inferred from an amplification in the carbon isotope excursion in soil organic matter (Bowen et al., 2004), but vegetation analysis inferred a decrease of ~40% in precipitation at the beginning of the PETM (Wing et al., 2005). Drier PETM conditions occurred probably in Utah (USA, paleolatitude ~45 °N; Bowen & Bowen, 2008); these findings are, however, controversial, since other studies (Retallack, 2005) suggest enhanced rainfall for this region (Bowen & Bowen, 2009; Retallack, 2009). Droughts by reduced soil moisture and biomass burning through wildfires in the subtropics during the PETM could have provided a significant carbon release to the atmosphere (Fig. 7).
In western Europe, sedimentary records from the Spanish Pyrenees (Schmitz & Pujalte, 2007) indicate seasonally increased precipitation during the PETM, leading to enhanced runoff into the Tethys Ocean and thus enhanced productivity by a rise in nutrient availability in the near-shore areas (Schmitz et al., 1996; Speijer & Wagner, 2002; Gavrilov et al., 2003). Increased precipitation over England (~+1 mm/day change from 4xCO2 to 16xCO2) would also have generated a feedback on the carbon cycle, for example enhanced carbon emission from wetlands (Pancost et al., 2007).
3. Feedbacks associated with the PETM ocean circulation
In this section, two feedback loops involving the carbon cycle and climate are discussed. The first is associated with the rise of greenhouse gas concentrations (water vapor, CO2, CH4, and other gases) in the atmosphere due to tectonic changes such as volcanism (Bralower et al., 1997; Kennett & Stott, 1991; Lyle et al., 2008; Sluijs et al., 2007; Storey et al., 2007; Svensen et al., 2004) and the second with the response of the climate system to regional or global sea level change by tectonic uplift and climatic changes (Fig. 2). Evidence of marine transgression during the PETM (Handley et al., 2011; Maclennan & Jones, 2006; Schmitz & Pujalte, 2003; Sluijs et al., 2008b) related to changes in spreading rate, volcanism, and regional perturbations as well as climatic changes (melting of glaciers, thermal expansion, and changes in ocean circulation) suggests that sea levels rose by approximately 20-30 m. The increase in surface temperatures and freshening of the sea surface by enhanced poleward moisture transport in response to a rise of greenhouse gases changes the regional buoyancy and momentum fluxes, leading to changes in vertical density gradients and stratification of the deep sea. Warmer and more saline subtropical water masses are modeled associated with the initial PETM warming (~0.2 psu higher in salinity for the 8xCO2 than for the 4xCO2 experiment), originating near the Gulf of Mexico and mixed via the eastern North Atlantic into intermediate layers. For intermediate water masses in the North Atlantic Ocean, simulated temperature rises by ~4°C (from ~11°C to ~15°C) (Fig. 8a).
In the Pacific, the increase of the atmospheric CO2 to 8xCO2 results in an increase in the vertical density gradient, since surface waters become significantly lighter with the warming. Deep-sea temperatures increase by ~2.5°C due to the global warming (Fig. 8b). The simulated Pacific circulation in the 4xCO2 scenario is nearly symmetric about the equator, with deep-sea ventilation occurring in the polar regions of the northern and southern hemisphere (Fig. 9a), in agreement with analyses of Nd isotope data that indicated a bimodal ventilation (D.J. Thomas et al., 2008). The northward-directed Atlantic deep-sea circulation of ~4 Sv (1 Sv = 106 m3 s-1) in the 4xCO2 scenario with a source of deep-water formation in the South Atlantic is comparable in strength with the modern but reversed.
With an increase of the CO2-radiative forcing to 8xCO2, the ventilation of the deep sea is reduced and the age of water masses in intermediate depth is increased (Fig. 9b), particularly in the southern high latitudes. The Atlantic deep-sea circulation in the 8xCO2 scenario remains reversed, in agreement with Zeebe & Zachos (2007), who used inferred [CO32-] gradients in the deep sea, but in contrast with the reconstruction of an abrupt shift in the deep-sea circulation during the PETM to a North Atlantic deep-water source, based on benthic carbon isotope records (Nunes & Norris, 2006).
The warming of intermediate and deep-water masses could have had a positive feedback on the ocean circulation (Fig. 2), as proposed in Bice & Marotzke (2002). The warming of the ocean by changes in the buoyancy forcing (heat and freshwater fluxes) and circulation lowers the depth of methane hydrate stability, which depends on pressure, temperature, salinity, and gas composition, from ~900 m to ~1500 m (Fig. 10; Dickens et al., 1995). This change might have triggered a massive methane hydrate release into the atmosphere-ocean system, which in turn accelerated the global warming (Archer & Buffett, 2005). The potential consequences of such an amplification are displayed in Figs. 8c and d, for the assumption that the carbon release corresponded to ~4400 PgC (16xCO2 experiment) relative to the 8xCO2 experiment. A temperature increase of >3.5˚C is simulated for high-latitude intermediate water masses in the Pacific due to an increase in vertical density gradients. The increase in the ideal age of water masses is shown in Fig. 9c.
The feedback loop associated with the sea level changes at the PETM would have affected the oceans by an enhanced freshing from Arctic Ocean. An increased flow via the Turgay Strait, the passage between the Arctic and Tethys Ocean, has been inferred from the abundance of dinoflagellate cysts (e.g. Iakokleva et al., 2001). Higher sea levels might also have allowed a throughflow via the Fram and Bering Straits, as supported by Nd-Sr isotopes in fish fossils (Gleason et al., 2009; Roberts et al., 2009), paleogeographic reconstructions (Scotese, 2011) and climate simulations (Cope & Winguth, 2011; Heinemann et al., 2009).
A freshwater input from the Arctic Ocean into the North Pacific Ocean (Marincovich & Gladenkov, 1999) would have produced an increase in the vertical density gradients and led to a weakening of the North Pacific intermediate water masses by 2.5 Sv at 30°N, and a comparable increase in the Pacific deep-sea circulation. The opening of the Bering Strait would have shifted formation of intermediate water masses in the North Pacific more equatorward towards the arid subtropics, by that increasing temperature and salinity of intermediate water masses (Fig. 11). Such a temperature change in intermediate waters at the beginning of the PETM warming could have contributed to the release of methane hydrates (e.g. Kennett & Stott, 1991; Sluijs et al., 2007). Most of the methane released from the hydrates would have ultimately reached the atmosphere or oxidized as CO2 and thus increased the greenhouse gas radiative forcing during the PETM (Fig. 2).
4. Feedbacks associated with the atmospheric chemistry during the PETM
Many potentially important feedback processes are associated with atmospheric chemistry (Beerling et al., 2007).
Possible changes associated with clouds at the beginning of the PETM are for example cloud albedo, cloud optical depth, or heat transport by tropical cyclones. Clouds interfere with the transfer of radiation because they reflect a certain amount of radiation back to space and they act as a blanket for thermal radiation. The reflectivity of clouds is influenced by cloud condensation nuclei (CCN). While today’s major source for CCN over land is due to pollution, CCN concentrations over remote ocean areas are linked to marine productivity via dimethyl sulfide (DMS) emission from the ocean. DMS emitted from certain phytoplankton groups is mixed into the troposphere and is oxidized to sulfate particles, which then act as CCN for marine clouds. The CCN concentration affects cloud droplet size and distribution, which influences cloud reflectivity and hence the climate. Climate change on the large scale, in turn, affects the ocean circulation, nutrient cycles and consequently the phytoplankton concentration in the oceans and thereby closes via DMS emission the feedback loop, as first hypothesized by Charlson et al. (1987). If global productivity had declined during the PETM by ocean stagnation and reduced equatorial upwelling, the concentration of CCN would also have been reduced.
This would have affected the cloud optical depth (Kump and Pollard, 2008), leading to high-latitude warming and a further increase in ocean stratification and stagnation of the deep-sea circulation, probably similar to the one modeled in the 16xCO2 experiment by Winguth et al. (2010). Polar stratospheric clouds (Sloan & Pollard, 1998; Kirk-Davidoff et al., 2002) or intensified tropical cyclone activity (Korty et al., 2008) could have further exaggerated warming at the PETM.
Another feedback between the carbon cycle and the climate that may have played an important role during the PETM are volatile organic compounds (VOCs; Beerling et al., 2007). VOCs are emitted by plants, for example isoprene with present-day emission rates comparable to that of methane (Guenther et al., 2006; Prather & Erhalt, 2001). Isoprene is a major player in the oxidative chemistry of the troposphere and influences the formation of tropospheric ozone (Fehsenfeld et al., 1992), decreases the hydroxyl radical concentration, increases the residence time of CH4, and is involved in forming organic aerosols influencing the climate by acting as CCN (Beerling et al., 2007).
High CH4 emissions during the PETM could have increased the atmospheric methane concentration and enhanced radiative forcing (with a ~21 times higher global warming potential than CO2 over a time span of 100 years; IPCC, 1990). Methane in the atmosphere is typically either reduced by oxidation to CO2 or interacts with other chemical components. Emission scenarios for the PETM considering atmospheric chemistry involving NOx and ozone reactions indicate that the life-time of methane in the atmosphere increases with increasing emission of methane, thus leading to an increased radiative forcing influencing the climate and methane hydrate destabilization within a positive feedback loop (Schmidt & Shindell, 2003).
5. Feedbacks associated with weathering during the PETM
While the feedbacks listed in the previous sections illustrate the complexity of the PETM warming, the rapid recovery phase after the CIE remains controversial as well. Rapidly regrowing organic carbon stocks on land and in the ocean on climatic time scales <104 years may have contributed to a draw-down of the atmospheric CO2 concentration (Bowen & Zachos, 2010), thus creating a positive feedback between a cooler climate and a more vigorous ocean circulation with reduced vertical density gradients and enhanced ventilation from high latitudes (Fig. 12). Intensification of wind-driven upwelling and enhanced high-latitude mixing stimulate global productivity through higher nutrient availability in the euphotic zone. Such an increase in the productivity (Bains et al., 1999; Stoll et al., 2007; Sluijs et al., 2006) could eventually have accelerated the draw-down of the atmospheric CO2.
On longer geologic timescales (>104 yrs), carbon sequestration by weathering of continental rocks becomes an important process. Atmospheric CO2 and H2O reacts with rocks and is converted into aqueous bicarbonate that is transferred to the oceans via riverine discharge and eventually deposited on the seafloor as biogenic carbonates (Walker et al., 1981; Berner, 2004).
The hothouse climate during the PETM with an increase in precipitation and plant growth likely accelerated weathering. The associated large input of dissolved bicarbonates into the ocean would have neutralized the oceans’ acidity and led to post-CIE deepening of the lysocline (Zachos et al., 2005), and preservation of calcareous marine sediments (Fig. 12; Kelly et al., 2010). This negative weathering feedback would ultimately have led to a draw-down of atmospheric CO2, climatic cooling and reduced weathering.
6. Conclusive remarks
The PETM, represented by the largest perturbation in climate and carbon cyle during the last 60 million years (Fig. 1; Pearson & Palmer, 2000; Royer et al., 2007) can be considered as an analog for future climate change. The analysis of ice bubbles trapped in the Antarctic suggests a variability of the atmospheric CO2 concentration over the last 800,000 yrs ranging from 172 ppmv to 300 ppmv for the preindustrial period. As a result of human activities, CO2 in the atmosphere rose over the last couple of hundred years with a pace not seen in recent geological history. In the year 2011, the atmospheric CO2 concentration exceeded 390 ppmv (Tans & Keeling, 2011), and a doubling of the pre-industrial atmospheric CO2 level is expected by the end of this century. The climate sensitivity for this doubling in CO2 is estimated to be 1.9–6.2 K due to the positive forcings, i.e. the rise in greenhouse gases, and including the negative forcing arising from the cooling effects of aerosols (IPCC, 2007; Andreae, 2007). A release of ~2000 PgC into the atmosphere in the next couple of hundred years could eventually trigger the release of an additional 2000-4000 PgC from marine sediments (Archer & Buffett, 2005), a flux comparable to that observed at the PETM (Zachos et al., 2008) and more than 10 times higher than observed during the last million years. The additional carbon release would act as a positive feedback, accelerating the warming.
The climatic and biogeochemical response to remarkable carbon emissions would likely be severe, for example a more frequent occurrence of climate extremes (heat waves, droughts and floods), particularly over the continents and at high latitudes, as well as ocean warming and stagnation. Another likely effect is ocean acidification and a rise of the calcite dissolution depth (Zachos et al., 2005), affecting marine organisms with calcareous shells (E. Thomas, 1998, 2003, 2007). The increased vertical gradients in the ocean together with warmer temperatures might produce near-anoxic conditions in the oxygen minimum zone (comparable with dead zones in the Black Sea or the Gulf of Mexico). Geochemical evidence for the PETM supports a downward expansion of the oxygen-minimum zone below 1500 m (Chun et al., 2010; Nicolo et al., 2011) in agreement with foraminiferal evidence.
All model simulations were done on NCAR computers, supported by NSF. The work is supported by NSF Grant EAR-0628336.