research objectives

The early Eocene hyperthermals represent natural carbon cycle experiments involving rapid input of carbon to the atmosphere/ocean on scales comparable to the modern anthropogenic forcing. The success of recent ODP expeditions in recovering spectacular quality and complete sections of these events in depth transects coupled with recent advances in the development of coupled earth system models provide new opportunities to address the critical questions framed below. This effort, however, will require a highly focused, systematic reevaluation of how ocean carbon chemistry changed during such events, specifically regarding the magnitude of the CIEs and variations in carbonate accumulation rates on a global scale. It will also require a carefully coordinated series of modeling experiments and integration of observational data.  Consequently the objectives of this project, framed as critical questions, are as follows:

 

1) What was the mass and rate of carbon released during the hyperthermals and was this release instantaneous or pulsed?

Detailed carbon isotope records exist for many marine and terrestrial PETM sections, and provide a key constraint for quantifying the mass and rate of carbon release to the ocean/atmosphere. However, two significant questions remain. First, the amplitude of the CIEs in pelagic, marine sections typically ranges from 2 to 4ä, while the CIE in terrestrial sections ranges from 5 to 6ä, so the true amplitude of carbon release remains uncertain.  Second, coccolith and bulk carbonate isotope records show steps in the negative isotope excursion which are not observed in planktonic foraminiferal (single specimen) records, so it is unclear whether carbon was released in a single instantaneous event or in multiple pulses that would reflect periodic forcing or positive feedbacks.  Resolution of both questions rests on identifying potential biological or vital effects on isotope fractionation which may bias components of marine or terrestrial records, and on assessing the degree to which variable dissolution of marine foraminifera and coccoliths may alter the preserved isotopic record.  For example, changing CO2 and humidity may amplify the CIE in terrestrial records (e.g., Bowen et al., 2004) while ocean acidification and lower [CO32-] of surface waters could diminish the marine CIE through a carbonate ion effect (e.g. Spero et al., 1997). An alternative hypothesis, that marine carbonate records were truncated by dissolution and sediment reworking, is consistent with depth transects where minimum CIE values recorded in bulk carbonate increases with depth (Kelly et al., 1996; Zachos et al., 2005).   A combination of dissolution and reworking could also potentially introduce steps in the marine carbon isotopic record, as could variable vital effects in coccolith isotopic fractionation.

To empirically address these issues, we need to identify and constrain potential artifacts associated with microfossil shell production and deposition. To start, accumulation rates must be more tightly constrained. Dissolution horizons should be associated with sharp spikes in the accumulation of extraterrestrial He (3HeET), and be more pronounced in deeper sites. Also, the lack of transitional d13C values in single planktonic shells might be an artifact of sampling biases, which could be resolved by analyzing rare taxa that might have displaced common taxa during the transitions (Kelly et al., 1998). A second solution is to measure the d13C of organic compounds. Recent investigations show the presence of n-alkanes derived from terrestrial plant leaf waxes in marine PETM sections (Hasegawa et al., submitted; Pagani et al., in press). The d13C n-alk will allow for a direct measure of the d13Cterr relative to the marine carbonate record. Another strategy involves targeting near-shore marine sections that are relatively expanded (Giusberti et al., submitted; Zachos et al., submitted). And finally, interpretation of both marine and terrestrial archives of d13CDIC requires application of models that can evaluate the parameters that influence carbon isotope fractionation (e.g., Bowen et al., 2004).  For this purpose, box models provide a fast and effective means of simulating the whole-ocean, time dependent C- isotope evolution over long time scales. However, circulation models that simulate the effects of deep-water aging and carbonate dissolution on sedimentation are clearly required to interpret the spatial variations recorded in bulk and benthic foraminiferal d13C records.

2) How do the magnitude and rate of carbon release during hyperthermals affect the scale of the changes in the lysocline and CaCO3 accumulation rates?

      This question can be addressed by distinguishing between the acidification phase when carbon was being introduced and CCD/lysocline were rapidly shoaling, and the recovery phase, when carbon was being buffered and the same surfaces were deepening.

a) Acidification phase; carbonate undersaturation and chemical erosion. As discussed above, all PETM and ELMO sections are characterized by carbonate dissolution horizons. While dissolution is more pronounced at deeper sites, as expected (Fig. 1), spatially, there are significant differences in carbonate deposition patterns that are difficult to reconcile with simple models of ocean acidification.  For example, in the Southern Ocean PETM intervals at Sites 689 and 680, the carbonate content drops, but only to 75 and 85% respectively, implying the local CCD did not shoal as much as elsewhere (Schellenberg et al., in prep).  To the north, in the equatorial Atlantic, Caribbean, and North Atlantic, some sites show distinct clay layers with little to no carbonate (Bralower et al., 1997), while others show less carbonate dissolution. One hypothesis suggests that these spatial variations reflect the release and entry path of carbon into the deep sea.  Alternatively, these trends may reflect the influence of circulation on local productivity, inorganic/organic carbon fluxes, and bottom water redox conditions (Ridgwell et al., 2005).  These factors must be delineated to infer the full magnitude of carbonate dissolution.

b) Recovery phase: mass /rate of carbonate deposited (chalk layer). During recovery, deep-sea [CO32-] transitions (in ~20-40 ky) from an extreme undersaturated state to what appears to be a supersaturated state as characterized by pure white, rapidly-deposited coccolith-rich chalks (Farley and Eltgroth, 2003; Kelly et al., 2005). The white chalk appears to be a common feature of most pelagic PETM sections. This response is not unexpected as carbon cycle models that utilize a sediment/rock weathering feedback exhibit this type of behavior. The oversaturation results directly from the loading of dissolved ions from accelerated dissolution of carbonates and silicates and, hence, reflects an important negative feedback.  As CO2 is extracted from the system during weathering and replaced with carbonate ion, [CO32-] rises forcing the carbonate saturation horizon to great depths, thus allowing the ocean to purge the excess ions. As a consequence, carbonate sediment accumulation rates should rise. The issue we wish to resolve is where, and to what extent?

The magnitude of carbonate accumulation and carbon sequestration during the recovery interval remains highly uncertain because contrasting age models imply significantly different rates.  Application of an orbital age model (Ršhl et al., 2000) to the white chalk at Site 690, yields a duration of 100 ky, and creates an inflection in the C-isotope curves, which if real would imply a relative decrease in the burial of reduced carbon.  In contrast, the 3HeET model (Farley and Eltgroth, 2003) implies a much faster carbonate accumulation rate and shorter duration (20 ky) for the recovery interval.  In addition, the 3HeET age model implies that the removal of carbon from the ocean-atmosphere system during this phase was nearly as rapid as its addition to the ocean-atmosphere system, <20 ky. From a purely theoretical point of view, we might anticipate a rapid increase in carbonate accumulation rates during the period of supersaturation as the lysocline descends to the seafloor. Yet, increased carbonate/organic carbon burial ratios would have worked against the rapid rise in d13C implied by the 3HeET age model, unless increased carbonate ballasting also elevated organic carbon burial fluxes (although the organic/inorganic burial ratio would also have to increase).  These apparent paradoxes require further investigation.

Here, astronomical stratigraphy and 3HeET will be used to improve the estimates of carbonate accumulation on a global scale for the PETM and ELMO. Moreover, dissolution indexes, such as foraminiferal fragmentation and % coarse fraction will be used to constrain the position of the lysocline relative to the CCD. Because the differences in carbonate accumulation patterns almost certainly reflect on the influence of circulation on local productivity, inorganic/organic carbon fluxes, and bottom water redox conditions (Ridgwell et al., 2005), ocean models are required to assess these affects under a variety of scenarios. Multiple simulations (sensitivity tests) will be required to gauge the potential effects of each variable using models that can simulate in 3-D each of critical controls on carbon fluxes and burial. 

 

3) How did lower/higher pH and other environmental changes affect production and/or calcification of coccolithophorids and foraminifera, and their relative contributions to fluxes?

In theory, the more rapid the rate of CO2 input to the atmosphere, the more extreme the changes in surface ocean pH and carbonate saturation state.  In the modern ocean, planktonic calcifiers and corals are potentially threatened by future increases in CO2 (e.g., Ridgwell and Zeebe, 2005). With the PETM and ELMO, the major groups of planktonic calcifiers, coccolithophorids and foraminifera, did not experience major extinction implying that surface ocean pH did not reach extinction levels. This would be consistent with the relatively long period (>10 ky) over which the several thousand Pg of carbon were released.  Still, significant changes in the relative abundances of coccoliths and planktonic foraminifera have been documented (Bralower, 2002; Kelly, 2002; Kelly et al., 1996). Some of the changes can be attributed to warming, increased stratification, and lower fertility. Did lower/higher pH have any noticeable impacts on the calcification rates of coccolithophorids and foraminifera?

To address this, a systematic quantitative assessment of coccolithophorids and foraminiferal assemblages, size, morphology, calcification rates, and growth rates is required.  Such an assessment will help elucidate the degree to which changes in carbonate accumulation rates reflect shifts in the production by calcifying organisms vs. changes in the dissolution or preservation of carbonate.  Trace metal indicators of growth rates and ocean chemistry, and isotope ratios will be measured on individual species at the level of shells and chambers. Models of biogenic calcification will be required to interpret these data. Ultimately, these data will lead to improvements in the way our numerical models calculate changes in rates of calcification in the surface ocean resulting from ocean acidification events.

 

4) Was the mass of carbon released during the PETM substantially greater than during ELMO? Did feedbacks play a role in amplifying the fluxes? 

Given the findings from Objectives 2 and 3, can we quantify the mass and rate of carbon input to the ocean during the hyperthermals?  Computations based solely on the CIE magnitude cannot provide a unique solution. An independent constraint is required. The degree of the model-simulated ocean acidification required to match reconstructed CCD shifts in the Atlantic and Pacific allows an independent estimate of total carbon input. The simulations will provide a temporal analysis of CCD variations, carbonate accumulation rates, d13C variations in surface and deep waters, and climate change, which, when constrained by observation, will permit assessment of magnitude, rate, and origin of the carbon released. Can the carbon flux from any one source (reservoir) account for the full mass of carbon added to the ocean/atmosphere? If not, what are the potential feedbacks that might supply additional carbon? Solution of the above questions is critical if we are to precisely compare the scale of the hyperthermals to the modern predicament.

 

5) How did feedback processes restore ocean carbonate chemistry and atmospheric CO2?

Although the focus of past research has been on the onset and magnitude of the CIE, it is equally important to understand the time scale and mechanisms of recovery of the biosphere to massive carbon addition if we are to gauge the long-term implications of fossil fuel burning.  Are there feedbacks that accelerate or damp this response?

Model sensitivity tests are required to simulate the time scale of C-cycle recovery from rapid injections of CO2. One component of this recovery is weathering of carbonate (short-term) and silicate (long-term) rocks on land. Using a suite of numerical models, we will investigate the sensitivity of CO2 consumption rates during weathering to temperature, freshwater fluxes, and soil development. Another important factor is the extent to which marine productivity and organic carbon burial are promoted or inhibited by changes in riverine nutrient delivery, climate, and ocean circulation associated with the hyperthermals. If the 3HeET age model is correct, an organic carbon burial event rivaling the carbon release itself must have occurred. What drove this response of the marine biota to the PETM and did a similar response occur during ELMO?

 

RESEARCH STRATEGY

Research Team

We have assembled an interdisciplinary research team with extensive experience in the core areas of this project. The project team is organized into two groups; a theoretical group (L. Kump (group leader), co-PIs G. Bowen, A. Winguth, and R. Zeebe, and international collaborators, A. Ridgwell and D. Beerling), and an observational group (J. Zachos (group leader), T. Bralower, K. Farley C. Kelly, M. Pagani, and H. Stoll, and international collaborators U. Ršhl, H. Brinkhuis, S. Gibbs, T. Tyrrell, I. Raffi, S. Galeotti, & P. Ziveri). We will continue to collaborate with investigators of the 2001 biocomplexity project (J. Dickens, P. Koch, E. Thomas, L. Sloan, P. Delaney, S. Schellenberg, C. Hollis, J. Kiehl), as well as several participants of ODP Legs 198, 207, and 208 (K. Kaiho, R. Norris, D. Kroon, L. Lourens, M. Petrizzo, M. Nicolo. S. Monechi).

 

Observational Database: Field and Lab Work

Sampling Network

To constrain and test the model experiments, it will be necessary to improve the temporal resolution and spatial coverage of reconstructions of ocean carbon chemistry and flux changes during the hyperthermals. This includes reconstruction of d13CDIC, biogenic carbonate production, fluxes, dissolution and burial. To this end, we will construct high-resolution records for each site in then N. Pacific and S. Atlantic ODP depth transects (Legs 198, 207, and 208), plus a dozen or more pelagic and hemipelagic sites so that each major ocean sector, the North Atlantic (e.g. Sites 401, 549, 929), Tethys (e.g. Piave, Italy, Zumaya, Spain), central Pacific (Sites 1220, 1221), Southern Ocean (Sites 690,738, Dee & Mead Stream), and Indian Ocean (e.g., Sites 213, 762), is well represented for both events at multiple depths.  Based on our preliminary survey, it appears ELMO is present at many of the same sites where the PETM has been documented (e.g.; Charisi & Schmitz, 1998; Speijer & Schmitz, 1998; Galeotti et al., 2000; Zachos et al., 2003; Crouch et al., 2001; Schmitz et al., 2001; Hancock et al., 2002; Egger et al., 2002; Cramer et al., 2003; Zachos, Kroon, Blum et al., 2003; Hancock et al., 2003; Crouch et al., 2003; McCarren et al., 2005; in prep., Hollis et al., 2005; Lourens et al., 2005; Nicolo et al., 2005; Nunes & Norris, 2006; Gusberti et al., submitted; Westerhold et al, 2005; in prep., Galeotti et al., in press; Kaiho et al., in press), so spatial coverage should be as good, if not better.

 

Astronomical / Extraterrestrial 3He Age Control

Highly precise time scales are fundamental to both the observational and modeling component of this project. New sections combined with recent improvements in techniques offer great potential to resolve chronology for PETM and ELMO sections.  Thus, substantial effort will go into refining age models and computing the carbonate accumulation rates between and across the two hyperthermals. With the exception of the clay layers, orbital stratigraphy coupled with chemostratigraphy will be used to establish baseline accumulation rates prior to and after the events. Ba concentration data, obtained from XRF scanning, offer the potential to resolve cycles where other elemental data are non-cyclic. The Leg 208 orbitally tuned time-series, which is being developed using a combination of the high-resolution (~1-3 cm) core logging, XRF core scanner and isotope data from all 5 sites, spans the entire interval from C24N to C25N and includes both the PETM and ELMO (Fig. 2), and will thus serve as the primary section to which all others are correlated above and below the clay layers.

In the clay or condensed layers, we propose to expand on earlier work to establish the pace of the PETM and ELMO using extraterrestrial 3He as a constant flux proxy. 3He in most deep-sea sediments is overwhelmingly derived from ~10 nm diameter interplanetary dust particles (IDPs). Because terrestrial and extraterrestrial helium have very different 3He/4He ratios, isotopic measurements can be used to establish the concentration of extraterrestrial 3He (3HeET) in a sample.  For a given 3HeET flux (F, in atoms/area/time), 3HeET is related to the sediment mass accumulation rate (a) by dilution: 3He=F/a. This relationship can be inverted to calculate mass accumulation rates from 3HeET measurements, provided F can be determined, for example by measuring the amount of 3HeET accumulated over an orbitally-tuned time interval. Temporal variability of the apparent 3He flux to Earth over intervals comparable to the PETM is fairly small. Uncertainties in the apparent flux can be reduced by calibration using a sedimentary section located close in space and time to the section of interest. Similarly, comparisons among multiple sites provide an empirical test of the resulting age models, because sedimentary phenomena affecting the apparent flux are unlikely to be identical in timing and magnitude at geographically separated localities.  The 3He method provides instantaneous sedimentation mass accumulation rate estimates for every analyzed sample, which can be converted into linear sedimentation rates, and by integration, used to establish an age model. This method successfully resolved fine scale changes in accumulation rate at ODP Sites 690 and 1051 (Farley and Eltgroth, 2003).  The method will be applied to each site in the depth transects, plus a half-dozen sites in the North Atlantic and Indian Ocean. Preliminary data for Site 1209 from the Leg 198 transect shows a 20-fold increase in concentrations (similar to clay) suggesting a substantial loss of carbonate (Fig. 3).

 

Global Carbon Isotope Excursions

Detailed, carbonate based carbon isotope records exist for almost every PETM boundary section studied to date. This includes bulk, planktonic and benthic records. In the Leg 198/208 depth transects these records are currently being extended upward to include ELMO (Fig. 2).  We will generate similar records for ELMO at a half-dozen additional sites where the PETM is present. The resolution will be on the order of 1-2 ky for the bulk carbon isotope records, and at lower resolution for planktonic and benthic foraminifera (2-4 ky).  Since there are discrepancies between coccolith-dominated bulk and planktonic foraminiferal stable isotope records from some sites during the PETM, we will also generate stable isotope records in restricted size fractions dominated by single coccolith genera, eliminating artifacts of changing species assemblages and in conjunction with cellular models elucidating the contribution of vital effects in stable isotope records (e.g., Stoll, 2005). Coccolith fraction stable isotope records will be produced in combination with trace element records from coccoliths which verify the primary rather than diagenetic nature of the coccolith calcite and reveal unique geochemical signatures in certain species during isotope plateaus which cannot be artifacts of mixing of pre-PETM and PETM material.

Understanding the full extent of the negative CIE during the PETM (and ELMO) is critical to modeling experiments (see above). Theoretically, a relatively larger terrestrial PETM CIE, measured from d13C values of terrestrial leaf wax n-alkanes in the Arctic region, could be potentially explained by a floral shift from gymnosperms to angiosperms.  However, similar floral shifts must have synchronously occurred at other localities to account for terrestrial 13C-excursions, derived from soil carbonates, of the same magnitude.  Rather than explaining a similar observation by different processes, it is possible that the CIE expressed by higher plant n-alkanes reflects the full d13C change of atmospheric CO2 in equilibrium with the surface ocean during the PETM.  This scenario implies that foraminiferal and coccolith d13C values do not accurately represent the full CIE of ocean DIC due to effects related to dissolution and changes in pH.  If this hypothesis is validated by the extensive compound-specific isotope analysis proposed in this study, evidence for a ~ –5 to -6ä CIE would require a substantial increase in the mass of carbon released.  To test this, we intend to measure the carbon isotope ratios of n-alkanes extracted from the near-shore PETM boundary sections. High-resolution planktonic foraminifer carbon isotope records already exist for two shallow marine sections from the New Jersey margin (Zachos et al., in press), and two from outer shelf, upper slope sections from the California margin (John et al., 2005). The California sections are currently being processed and preliminary results indicate abundant terrestrially derived n-alkanes are present.  We will also sample hemipelagic sections along the Tethyan margin (e.g., Piave, Italy; Giusberti et al., submitted), and in New Zealand (Mead and Dee Stream; Nicolo et al., 2005), and several pelagic cores), to isolate n-alkanes.  Moreover, similar coupled n-alkane d13C and carbonate d13C records for ELMO when carbon flux was smaller and marine carbonate dissolution less severe, should provide additional insight into the nature of the offset. 

 

Biogenic Carbonate Accumulation

 High-resolution %CaCO3 records already exist for most of the pelagic and hemipelagic PETM sections in our global sampling network. Moreover, all the depth transect cores have also been scanned by XRF, and thus have high-resolution Ca concentration records (U. Ršhl, pers. comm.), which can be easily transformed through regression to %CaCO3. Thus, most of our analytical effort will focus on older DSDP and ODP cores and land-based outcrops.  The archive halves of cores that have not been scanned will be shipped to Bremen University for scanning. For the few cores that are unsuitable for scanning, discrete samples will be collected for conventional coulometric analysis. For most land-based marine PETM and some ELMO sections, CaCO3 records already exist, so the number of sections to resample will be minimal.

The relative contribution of foraminifera vs. coccolithophorids to the carbonate flux shifts significantly through the PETM (Fig. 4). To ascertain the global extent of this important ecological shift, standard sieving and settling techniques will be used to partition and compute the accumulation of various biogenic components (forams vs. coccoliths). Samples from the depth transects have already been wet sieved at >38 µm and dry sieved in larger size fractions (>63, >150 µm). Settling techniques will be used to separate nannofossils (Stoll, 2005; Stoll and Ziveri, 2002).

 

Dissolution CCD/Lysocline:

In sites examined to date, the foraminiferal preservation in the white chalk/ooze layer is excellent, with little to no fragmentation (Kelly et al., 2005; in prep.). In the Leg 208 depth transect, %CF (& %CaCO3) in this layer is essentially identical at all depths (Fig. 4), an observation that suggests the entire water column was highly oversaturated. To determine the scale of this unusual phenomenon, we will construct proxy preservation records for both the PETM and ELMO in each ocean sector (as above). We will target sites that provide the broadest depth coverage possible.  Proxies to be measured include the %CF, planktonic/benthic and fragmentation indexes, and microscale carbonate overgrowth. In the Leg 198 and 208 sites, we will investigate proxies sensitive to carbonate ion content such as depth-dependent variations in Zn/Ca in benthic foraminifera and Mg/Ca in planktonic foraminifera (Fehrenbacher and Martin, 2006).

Nannofossil preservation in PETM and ELMO samples will be evaluated by establishing the dissolution susceptibility of key nannofossil and foraminiferal species. This will be performed in three different ways:  (1) To establish the solution susceptibility of ancient plankton taxa, samples will be reacted with seawater that is undersaturated with respect to CaCO3, as done by Hill (1975) with traps. We will conduct similar experiments, but in the laboratory using seawater solutions of controlled pH. (2) We are compiling a global dataset of PETM nannofossil and foraminiferal assemblage data and will use multivariate techniques (Detrended Correspondence Analysis (DCA)) to determine ecological groupings of taxa (Gibbs, Bralower, et al., in prep.). This same data set can be mined to determine the relative amount of dissolution of individual samples as well as the solution susceptibility of different taxa. And (3) We will quantify % etched or overgrown taxa for nannofossil and foraminiferal specimens and use quantitative metrics to determine the extent of alteration using a scanning electron microscope. 

For planktonics, shell diameter/mass of specimens is also a useful gauge of their preservation (Bijma et al., 2002; Broecker, 2002).  Average mass (and wall thickness) of globally distributed mixed-layer planktonic species Acarinina soldadoesnsis, and thermocline dweller Subbotina patagonica will be measured. In addition, we will measure the weight of Morozovella velascoensis and the Òexcursion taxaÓ of Morozovella (M. allisonensis) and Acarinina (A. sibaiyensis) that are restricted to tropical and temperate locations e.g., (Kelly et al., 1998). Preliminary work shows a substantial increase in the mean mass of foraminiferal shells in the white, chalk layer. The fragmentation of foraminiferal assemblages will also be determined (e.g., (Kelly, 2002); Colosimo et al., 2005).

Once the relative solution susceptibility of calcareous nannofossil taxa is known, we will calibrate assemblage data with grain size analysis of the <38µm fraction. Grain size analysis offers an advantage for determining assemblage and preservational changes because it is fast and relatively inexpensive and provide us with the means to observe changes at unusually high resolution (mm to cm) with a minimum of effort. We are currently using a Malvern Mastersizer and Coulter Counter particle analyzer for this purpose.

 

Productivity of coccolithophore carbonate producers

To ascertain the extent to which changes in productivity of coccolithophorid algae contributed to shifts in the carbon cycle and carbonate accumulation during the PETM and ELMO, we will use trace element ratios in coccoliths representative of primary trace-metal incorporation during calcification.   The ratio of Sr/Ca in particular has been shown to vary with nutrient-stimulated growth rates in culture (Rickaby et al., 2002; Stoll et al., in review-a), sediment trap, and surface sediment transects (Stoll and Ziveri, 2004).  Where dissolution is minor, coccolith Sr/Ca ratios covary with other indicators of coccolithophorid production including alkenone and carbonate accumulation rates (Rickaby et al., in review).  This relationship likely arises due to variable polysaccharide extrusion in response to nutrient availability, which binds Sr preferentially to Ca and affects the near cell ratio of Sr/Ca available for uptake into the coccolith calcification vesicle (Langer et al., 2006). 

Coccolith Sr/Ca and other elemental ratios can be measured using separated near-monogeneric size fractions (e.g., Stoll and Bains, 2003), as well as a new ion probe analysis of individually picked populations of monogeneric coccoliths (Stoll et al., in review-b). This latter technique is especially advantageous for studying lower Eocene sections where coccoliths are not abundant due to dissolution, when monogeneric coccolith fractions cannot be satisfactorily separated from sediments, and in shelf sections where high lithogenic input may hamper ICP analysis of coccolith size fractions.  The ion probe technique also provides a clearer picture of how ecologically distinct coccolith genera responded to the events. The ion probe technique reproduced to within 1% the Sr/Ca ratios of monospecific cultured Helicosphaera carteri samples which had also been measured by inductively coupled plasma atomic emission spectroscopy.  Analyses of replicate populations of Toweius coccoliths, from the same sample picked and analyzed several months apart, yield Sr/Ca ratios which differ by 1%.

 

Figure 5. Models and model co-ordination used to study the carbon cycle during the early Paleogene hyperthermal events. A) Three modeling approaches will be used, each involving specific opportunities and compromises in terms of the time scales and scope of processes that can be modeled. B) Modeling experiments (gray arrows) will be coordinated to assess each motivating question. Earth system models (GENIE and CCSM) will provide boundary conditions for the process-oriented models. Process model simulations will be designed to investigate problems identified during data/model validation of the Earth system models and to develop hypotheses to be tested with model simulations

Numerical Modeling

Our modeling efforts to date have begun to constrain the magnitude and duration of carbon release required to generate the observed PETM isotopic and sedimentary response (see below), but a number of outstanding questions remain and new issues of CIE magnitude, duration, and truncation have arisen (described above) that we must address. Thus, we propose simulations with well-tested community models of various degrees of complexity to predict the oceanic and sedimentary response to a range of carbon addition scenarios (including the first quantitative analysis of the ELMO event).  In addition we will  assess which of the scenarios (e.g., clathrate or thermogenic methane vs. fossil carbon – CO2 input, fast vs. slow input) generate the proper isotopic, climatic, and CCD responses (both regional and global) for the onset, magnitude, and recovery from these events. Transient simulations will be carried out with the intermediate complexity Earth system model GENIE. This modeling will be supported by higher-resolution time-slice simulations using the comprehensive climate model CCSM-3. CCSM-3 will provide improved simulations of ocean biogeochemistry for the initiation and termination of the PETM and ELMO events, and necessary boundary conditions to GENIE. In this work we benefit greatly from the international team of experts whose primary task is model validation for modern-day and future prediction. In addition, box models of carbon cycle processes will be used to conduct long (multimillion-year) simulations, and process-specific models will be used to address particular components of the global response. These modeling efforts will be coordinated: GENIE and CCSM-3 will provide climate and ocean transport rates to the box models, and the process-oriented and box models will guide improvements to the parameterizations of the GCMs (Fig. 5).

All spatially resolved models will use late Paleocene - early Eocene paleogeography and physiography derived from GIS layers assembled by C. R. Scotese (www.scotese.com) and the techniques of Markwick and Valdes (2004). To assess the sensitivity to these boundary conditions, the paleogeography and paleobathymetry produced by the Paleogeographic Atlas Project of the University of Chicago (presently used in GENIE modeling as well as in the work of Bice and Marotzke, 2001, 2002) will also be employed.

 

Process-oriented models

Long-term simulations of the carbon cycle spanning the PETM and ELMO will be conducted with global atmosphere-ocean-sediment carbon cycle models, building upon the initial results for the PETM of Dickens (1997). The ocean component resolves the different ocean basins, high and low-latitudes, and surface/intermediate/deep waters (Walker and Kasting, 1992; Dickens, 1997; Sigman et al., 1998; Zeebe and Archer, 2005); water fluxes are derived from the Earth system models.  Coupling to sediment models (Keir, 1982; Sundquist, 1986; Sigman et al., 1998) and inclusion of carbonate and silicate weathering feedbacks will allow quantitative analyses of carbon fluxes on time scales from thousands to millions of years. Model tracers include nutrients such as PO4, stable carbon isotopes (d13C), and total CO2 and total alkalinity from which atmospheric CO2 is calculated. These simulations will provide the baseline reservoir properties for initiating GENIE and CCSM-3 simulations.

A depth-explicit soil organic carbon and CO2 model including stable isotopes (Bowen and Beerling, 2004) will be used to explore soil carbon processes and carbon isotope ratios to be compared to paleosol records, at vertical resolutions that cannot be achieved with the coupled Earth Systems models.  Steady-state and dynamic vertical soil organic carbon decay and soil gas CO2 profiles can be simulated from derived soil physical property and organic carbon input and decay parameters. Additional constraints for the model simulations will be given by n-alkane carbon isotope data made available by this project. Updates to the model will include development of schemes for interfacing the soils model with GENIE and CCSM-3 for grid-level simulation of PETM and ELMO soils. 

.

Earth system model of intermediate complexity: The GENIE model

Intermediate complexity models present an unprecedented opportunity to perform time continuous simulations of geologic events like the PETM and ELMO (or both simultaneously); previous modeling has been largely restricted to snapshots using GCMs or 0-D box model simulations.  GENIE (www.genie.ac.uk) consists of interconnected modules treating atmosphere, ocean, sediments, ice sheets, land surface and vegetation, soils, sea ice, and ocean biogeochemistry. The model uses a highly efficient frictional geostrophic 3-D representation of the ocean, a 2-D energy and moisture balance atmosphere (after Weaver et al., 2001; coarse-resolution 3-D treatments are possible), and a dynamic and thermodynamic sea-ice model (Annan et al., 2005; Edwards and Marsh, 2005). A sediment component of GENIE is currently under construction by A. Ridgwell (see letter of support) and based in part on the work of Archer et al. (2002). This component calculates sediment accumulation rates (carbonate, non-carbonate, and organic carbon), sediment compositions (e.g., %CaCO3 and %Corg), and isotopic compositions. As such, it allows for an assessment of the extent to which the magnitude of the marine CIE has been truncated by erosion, as hypothesized above based on the higher-plant alkane d13C record. Kump and his graduate student, Karla Panchuk, together with Ridgwell, have been modifying GENIE for application to the PETM event. The first simulations of the PETM event are encouraging (Fig. 6). We are able to reproduce the Walvis Ridge carbonate record with a 6000 Pg C addition over 15ky, including the more rapid recovery of carbonate sedimentation at shallower depths. We propose here to evaluate the degree of the model-simulated ocean acidification required to match reconstructed CCD shifts in both Atlantic and Pacific sectors. Constrained by new estimates of carbonate accumulation rates and d13C chronologies, the modeling will permit assessment of magnitude, rate, and origin of the carbon release. Also, with d13C tracers, we can simulate bottom water gradients as captured in benthic foraminifera (Nunes & Norris, 2006), while including the effects of dissolution on sedimentary profiles.

Presently GENIE does not allow weathering responses to climate change, a critical part of the response to large carbon additions. Increased fluxes of kaolinite from the continents during the PETM (Gibson et al., 2000; Robert and Kennett, 1994), together with an Os isotope excursion (Ravizza et al., 2001) have been argued to reflect intensified silicate weathering, and thus may signal increased consumption of CO2 through chemical weathering (cf. Thiry, 2000). Thus we need to develop a weathering module for GENIE. The model will use a set of predictive functions to estimate clay mineral abundances in soils based on climate, vegetation type, bedrock mineralogy, and physiography, derived through multiple regression analysis of digital spatial data sets for modern soils (FAO, 2003) in the context of global climatological datasets (New et al., 2000), biome distributions (http://edcsns17.cr.usgs.gov/glcc/ globe_int.html), and physiography (USGS, 1996) in unglaciated, non-agricultural regions.  Early Paleogene soil mineralogy will be predicted using these same variables, which are explicitly specified or calculated in GENIE. A paleogeologic map for the Eocene (following Bluth and Kump, 1991) will be constructed as a senior-thesis project in the first year of funding. Weathering rates will be determined using empirical rate laws derived from the data on river loads from modern-day, monolithologic watersheds (e.g., Bluth and Kump, 1994; Gibbs et al., 1999). The effect of relief and soil cover will be parameterized by analogy with modern drainages (e.g., Milliman and Syvitski, 1992; Summerfield and Hulton, 1994; Hooke, 2000). Validation will include comparison of chemical erosion rates calculated for the major rivers of the world compared to observation for the modern; fair agreement has been achieved in the past (Gibbs and Kump, 1994; Ludwig et al., 1999).

Moreover, the organic carbon weathering and deposition components of GENIE have not been implemented. As described above, the organic carbon cycle seems to be an important part of the recovery from the PETM, one that has been neglected to date. We will work with Ridgwell to activate this part of the model.

As described above, one explanation for the smaller amplitude CIE in the marine vs. terrestrial records is the dependence of the d13C of marine carbonate on carbonate ion concentration. To evaluate this hypothesis, the carbonate ion effect will be incorporated into the carbon cycle models and the results compared with those generated without the effect.

Presently, GENIE does not resolve continental shelves and thus cannot fully represent the response to the PETM/ELMO perturbations. We will evaluate the sensitivity of CCD response to this partitioning by attaching low-, mid-, and high-latitude shelf ÒboxesÓ to the box model and to GENIE, with specified partitioning based on paleogeographic reconstructions of latitudinally dependent flooded shelf areas for the early Paleogene and estimated rates of carbonate deposition based on adjacent open-ocean water chemistries (after Opdyke and Wilkinson, 1993) from GENIE.

 

Time-slice experiments for onset and recovery of the PETM and ELMO with the Community Climate System Model (CCSM)

The CCSM-3 (see letter of collaboration) is a coupled climate model for simulating EarthÕs climate system (http://www.ccsm.ucar.edu/models/ccsm3.0/). Time-slice simulations with the CCSM3 for the onset and recovery of the CIE near the PETM and ELMO will be guided by previous simulations with older versions of the CCSM (e.g. (Huber et al., 2002; Huber and Caballero, 2003; Huber and Sloan, 2001; Shellito et al., 2003)) and coordinated with CCSM Paleoclimate Working Group. We are planning to carry our simulations of various greenhouse gas levels (i.e. 1xCO2, 2xCO2, and 8xCO2) to explore interactions of the climate and the carbon cycle, in particular how the carbon cycle responds to changes in the climate state. The output of these simulations will be used for model-data validation and input for the offline three-dimensional carbon cycle modeling studies as well as for the process-oriented models and GENIE (surface wind velocities and shear stresses). Moreover, the role of geography will be investigated by comparison with a modern reference simulation.

Output from CCSM-3 will be used as input for offline carbon cycle models in order to investigate effects of different ocean circulation patterns on water column processes, fluxes at the sediment/water interface, and sedimentary composition. The offline carbon cycle model has been developed at the University of Chicago and includes the prediction of water mass tracers based on the HAMOCC model (Maier-Reimer, 1993; Archer and Maier-Reimer, 1994; Archer et al., 2000) with prognostic macro- and micronutrients (P, Si, Fe), oxygen, carbon tracers (d13C), and a model ("MudsÓ) for oxic and anoxic diagenesis of shallow and deep-sea sediments based on an efficient solver for steady-state pore water and solid phase diffusion/reaction/advection equations (Archer et al., 2002). The sediment model used in the proposed project is similar to the one in GENIE and will be updated in collaboration with D. Archer. The dependence of carbon export rates on micronutrients (Fe) and mineral ballast (Armstrong et al., 2002) have been tested with the HAMOCC model and compared with carbon tracers and sediment traps (Howard et al., in press). We will improve a simplified sulfur cycle currently developed at the MPI Hamburg (E. Maier-Reimer) by adopting the code developed for GENIE. The off-line coupling of the carbon cycle model with POP_TM is under development and tested with ideal age tracers in collaboration with the NCAR Oceanography Section (OS) (S. Yeager, W. Large) and Climate Change Research Section (CCR) (J. Kiehl).

We will carry out experiments with both GENIE and CCSM-3 to study the sensitivity of the late Paleocene/early Eocene thermohaline circulation and carbon cycle to changes in the hydrological cycle, following the work of Bice and Marotzke (2001) that demonstrated a high degree of sensitivity of the ocean circulation to the amounts and location of freshwater discharge to the ocean. Possible switches in the thermohaline circulation related to changes in atmospheric moisture transport could lead to sufficient warming to destabilize seafloor gas hydrates over most of the world ocean to a water depth of at least 1900 m. We will determine changes in clathrate stability following the lead of Bice and Marotzke (2002) and calculate methane fluxes accordingly, but explicitly determining the delay in response resulting from the requirement for the warming to penetrate through the sedimentary pore waters to the clathrates. Furthermore, we will perform simulations similar to those of Heinze (2004) to address how significant ecological changes at the PETM (see Kelly, 2002, and Kelly et al., 2005, and references therein) likely altered POC fluxes into the abyssal layers and thus influenced the CCD. Sensitivity of the CCD to mineral ballasting of POC (Armstrong et al., 2002; Klaas and Archer, 2002; Howard et al., in press) will be also explored. Finally, we will continue our research on the effect that bioturbation on the capacity of seafloor dissolution to damp the lysocline response to CO2 (Ridgwell et al., 2005).