The Source of Odd-Carbon Chain N-Alkanes

Supplementary Information

Arctic’s hydrology during global warming at the Palaeocene-Eocene thermal maximum. Mark Pagani, Nikolai Pedentchouk, Matthew Huber,Appy Sluijs, Stefan Schouten, Henk Brinkhuis, Jaap S. Sinninghe Damsté, Gerald R. Dickens, and the IODP Expedition 302 Expedition Scientists.

The source of odd-carbon chain n-alkanes

It is generally accepted that n-alkanes in the range of C25 to C33, exhibiting a strong odd-over-even carbon chain predominance, are derived from higher plant leaf waxes1-6. However, a single study (Zegouagh et al., 1998)7 of modern Arctic sediments from the Laptev Sea challenges this common interpretation, and instead suggests these compounds are of algal origin, rather than higher plants. This study is at odds with our interpretation that the long-chain n-alkanes with a strong odd-over-even carbon number predominance in the ACEX core are derived from terrestrial plants from the continent, and if valid, challenges our interpretations of isotopic trends.

Zegouagh et al. (1998)7 interprets the origin of n-alkanes from only three surface sediments collected in the delta of the Lena River and adjacent areas of the Laptev Sea. They find long-chain n-alkanes in relatively low abundance with an odd-over-even predominance (with a decreasing carbon preference index from the delta to the sea). They conclude that these compounds derive from algae on the basis of several criteria:

(1) No evidence for macromolecular organic matter derived from higher plants.

The absence of pyrolysis products, such as phenols, characteristic of lignins and cutins from terrestrial sources, is one criteria used to argue against a terrestrial origin for long-chain n-alkanes from the Laptev Sea. However, leaf wax lipids in regions distant from terrestrial vegetation are primarily transported by aeolian transport; a process that excludes the bulk of plant tissue. Therefore, the absence of macromolecular organic matter cannot be used as proof as evidence against higher plant sources for long-chain n-alkanes.

(2) “Absence” of long-chain fatty acids of terrestrial origin and selective degradation of fatty acids

The hydrocarbon fraction of the Laptev Sea samples comprise a small amount of the total lipid extract (0.02%), suggesting very small contributions of n-alkanes for these sediments (i.e., Table 1 of Zegouagh et al., 1998)7. In a typical leaf, n-alkanes are always present and represent 10-90% of the lipid fraction. This is not the case for fatty acids, where in some instances long-chain fatty acids are absent (Fig. S1). Thus, the absence of long chain, even-carbon number fatty acids is not necessarily indicative of an absence of terrestrial input. Given that fatty acids in the C14-C22 represent the majority of the acid fraction in sediments analyzed from the Laptev Sea8, and the fact that most marine organisms do not make saturated hydrocarbons, we would expect to find a lipid profile similar to that described in Zegouagh et al. (1998)7for a system dominated by algal production and aeolian transported n-alkanes. That is, we would anticipate that molecular distributions would be dominated by low molecular weight fatty acids with a small contribution of n-alkanes with terrestrial characteristics.

Importantly, Zegouagh et al. (1998)7 use the absence of even-carbon numbered fatty acids in the range of C24 and C34 as the primary evidence for a non-terrestrial origin for their n-alkanes. However, their absence cannot be used as evidence against terrestrial input, but the presence of even-carbon number fatty acidshelps confirm the existence of terrestrial input. In contrast to the findings reported by Zegouagh et al. (1998)7, the most dominant compounds in the ACEX PETM sediments are even-carbon numbered fatty acids ranging from C24 to C34 with the C28 homologue dominating. This provides clear and supporting evidence that our PETM n-alkanes are terrestrial derived.

(3) Long chain n-alkanes in cultures of marine algae

Zegouagh et al. (1998)7 acknowledge that most algae produce short chain n-alkanes (e.g., n-C17). They also note that the presence of long chain n-alkanes has been observed in some algal species, but with no odd-over-even predominance. Zegouagh et al. (1998)7 cite two articles that suggest that some algae produce long-chain n-alkanes with an odd-over-even predominance9,10. However, the observations in these cited studies have never been confirmed and are likely due to contamination11. For example, a set of >100 unialgal cultures of marine diatoms12 have been studied and long-chain n-alkanes with a strong odd-carbon number predominance have never been observed in any culture (Sinninghe Damsté et al., unpublished results). In conclusion, the scientific basis for the interpretation that (some) marine algae produce long-chain n-alkanes with a strong odd-carbon number predominance is at best weak, and perhaps not valid.

(4) The possibility that long-chain odd n-alkenes from the freshwater algae B. braunii are the source of the long chain n-alkanes

Zegouagh et al. (1998)7 suggest that the source of odd-long-chain n-alkanes derives from diagenetic reduction of alkenes, potentially produced by B. braunii.Although B. braunii produces n-alkenes, we stress that B. braunii is a freshwater algae. The pathway of alkene reduction is not completely understood. Polyunsaturated alkenes are labile and rarely preserved in modern sediments. Reduction of alkenescould result in the production of alkanes or competed degraded. Therefore, it is possible that the reduction of alkenes could lead to the profile observed in the Laptev Sea7. However, this supposition is highly speculative and cannot be substantiated.

(5) Carbon isotopic compositions suggest long-chain n-alkanes derive from a marine source.

C3 plants have a fairly wide range of isotopic values, ranging from -20‰ to -35‰. More negative isotopic values occur in closed canopies where the 13C of atmospheric CO2 is relatively depleted in 13C due to respiration of organic carbon. C3 plants enriched in 13C are typically found in more water-stressed environments. Further, n-alkanes will be more depleted in 13C relative to bulk leaf by ~3 to 5‰. Most n-alkanes measured in the Zegouagh et al. (1998)7 article are in the range of -28 to -32‰. However, lower molecular n-alkanes (in the range of C16 to C24) were measured in only one sample. In general, high-molecular weight n-alkanes typical of terrestrial plants have 13C values very similar to those reported in Zegouagh et al. (1998)7. In a study of organic carbon and n-alkanes from surficial sediments and suspended matter in the Arctic Ob and Yenisei Rivers, 13C values of n-alkanes from terrestrial sources averaged -26.8‰ and -28.1‰, respectively13. These isotopic values parallel the results of Zegouagh et al. (1998)7 and strongly support an interpretation that the n-alkanes in the Laptev Sea are terrestrial, and not algal, in origin. Indeed, the study of Arctic Ob and Yenisei River sediments directly apply n-alkanes with odd-over-even predominance as a proxy for terrestrial plant (leaf wax) input for these Arctic sediments. This is a typical proxy approach in studies devoted to carbon cycling in the Arctic Ocean14.

It is interesting to note that the short-chain n-alkanes of presumed algal origin (C16 - C18 with C18 dominant) in Zegouagh et al. (1998)7 also have rather 13C-depleted signatures (-26 to -30‰). Zegouagh et al. (1998)7 argue that isotopic similarities between short- and long-chain n-alkanes support an interpretation of a common source. However, these short-chain n-alkanes can have both algal and bacterial sources15,16, with even-chain length n-alkanes in the range of n-C14 to n-C22 generally attributed to bacteria17. Several explanations can account for the depleted signatures of these low molecular weight n-alkanes, including algae growth under elevated CO2 (due to cold water temperatures and greater CO2 saturation), bacterial components utilizing isotopically depleted inorganic carbon (grown at depth in the water column), or bacterial reworking of higher molecular weight n-alkanes6,18.

In conclusion, the evidence and arguments presented by Zegouagh et al. (1998)7 for an algal origin for long-chain n-alkanes with an odd-over-even predominance are not robust. Further, n-alkanes from the Palaeogene Arctic have substantially higher carbon preference indices (higher values increasingly support terrestrial origins) than those reported in Zegouagh et al. (1998)7. Finally, D and 13C values of the PETM short-chain and long-chain n-alkanes are substantially different from each other and show a different evolution with time–strong evidence that they derive from different sources.

The character of sediments from the disturbed top of Core 32X

The collection of data measured for the top ~50 cm of Core 32Xreveals complex sedimentary characteristics due to severe drilling disturbance (Fig. S2). One relatively negative 13CTOC value coupled with abundant Apectodinium spp. suggests that this interval represents PETM-aged material. However, the top two samples of the 13CTOC curve indicate an influence of late Palaeocene strata. Samples evaluated horizontally across the core, but remaining at the same depth, give contradictory results. These aspects indicate that drilling disturbance resulted in an inhomogeneous mixture of late Palaeocene and PETM material at the top of Core 32X. In this regard, the leaf wax n-alkane isotope data are interesting. For example, 13C values of n-alkanes (13CnC27/29) gradually become more negative upcore, confirming PETM material. Further, hydrogen isotopes (DnC27/29) show the most D-enriched values of the whole succession in this interval. If these values also represent a mixture of pre-PETM and PETM material, this suggests that the 13C and D values of leaf waxes from the original lowest part of the PETM are likely more 13C-depleted and D-enriched, respectively.

Hydrological Cycle: Model-Data Comparison

To develop a more complete interpretation of the data presented in this study and that of Sluijs et al. (2006)19, we integrate model results and proxy data interpretations of hydrological state for the Palaeocene and earliest Eocene (or PETM where available). Despite the acknowledged weaknesses of models in reproducing early Palaeogene temperatures, we believe that analysis of climate model results and comparison with proxies is a useful process because it provides a physical framework for proxy interpretation where proxy constraints are missing or open to multiple interpretations. In this section we provide a climate model context for our interpretation of proxies for the hydrological cycle and demonstrate that the model compares well with data. We begin with a theoretical framework.

Theory: To determine whether an increase in the hydrological cycle occurred using paleoclimate proxies, the location of proxy sites within the large-scale evaporation (E) and precipitation (P) distribution is necessary. A ‘more intense’ hydrological cycle as commonly defined in atmospheric dynamics20 refers to increased water vapor transport from low to high latitudes, or in other words, a stronger divergence of water vapor fluxes. In steady state, stronger divergence requires an intensification of E in net E zones and a counterbalancing increase in P in net P zones, leading to an increased meridional gradient of E-P21. In steady state, provided that E>P equatorward of the subtropical margins (e.g., the work of Ziegler et al. (2004)22 suggests this has been the case since the Permian) there must be net P averaged over the extratropics. Thus, we expect to find an apparent drying of the subtropical regions to be consistent with an increase in the vigor of the hydrological cycle with a compensating moistening in high latitudes.

The issue that determines the high-latitude water balance is not a local argument (i.e., that increases in temperature must lead to an increase in evaporation; a true statement) but a global argument: that as long as high latitudes have cooler temperatures than the subtropics, we expect the source of water that drives the atmosphere’s hydrological cycle to be from the low latitudes and the sink to be in the high latitudes. This argument is most accurate in a zonal average sense, but is likely to apply generally because zonal advection of water vapor is rapid and maintains a fairly homogenous water vapor distribution in the zonal direction23.

Model and experiments: The results provided below are from fully coupled climate model simulations performed with boundary conditions suitable for the early Palaeogene. The model is the Community Climate System Model v. 1.4 developed at the National Center for Atmospheric Research24. The boundary conditions (topography, bathymetry, vegetation) and initial conditions are described in Sewall et al. (2000)25 Huber and Sloan (2001)26, and Huber et al. (2003)27. We show results from two simulations: one with specified carbon dioxide concentrations set at 1120 ppm (denoted PETM-like) and another at 560 ppm (denoted pre-PETM). The pre-PETM simulation’s ocean circulation is described in Huber et al., (2003; 2004)27,28 and the tropical ocean behavior is described in Huber and Caballero (2003)29. The PETM-like simulation has not been previously described in depth—it is qualitatively similar to the pre-PETM simulation albeit with substantial high-latitude and deep-ocean warming (deep-ocean and high-latitude oceans are approximately 11°C warmer than modern values).

Given theoretical constraints, we expecthigh-latitude soils to have been relatively close to saturation at all times, with increases in precipitation largely balanced by increases in runoff. This can easily be seen, since, in steady state:

E-P = R

where R is runoff. Provided that P>E as warming occurs, we expect increases in P to be balanced by changes in R. Soil moisture plays the role of a small storage term in the non-steady state balance, but does not appear in the steady-state balance other than implicitly through its potential feedbacks on E or P.

E-P, Relative Humidity, and Soil Moisture Distribution Results: Comparison of pre-PETM to PETM climate model simulations indicates a robust evaporation minus precipitation (E-P) global pattern with minor-to-no shifts in the regions experiencing net P or net E (Fig. S3). In both simulations, higher latitudes experience strongly net P, whereas low-to mid-latitudes experience strongly net E. One might argue that high-latitude evaporation is not strong enough in these simulations because of the model’s cool bias30. While this is likely true to some extent, we argue that the appearance of P>E at high latitudes is not simply an artifact of the cool bias of the model, but more of a reflection of the basic physical balance previously described. Comparison with fixed SST simulations, with SSTs specified to be close to proxy estimates for the PETM (thus obviating part of the cool bias), yield very similar E-P distributions. This robust spatial E-P pattern is consistent with proxy records22. In detail, the E-P pattern becomes more complex in regions strongly affected by topography such as the western interior of North America, where a orographically driven ‘rain shadow’ is adjacent to a monsoonal plume of precipitation that emerges from the Gulf of Mexico as a summer season low-level jet25. Although this variability is expected in mid-latitude regions with strong topographic variation, we anticipate the Arctic region to be a less complex environment within which the hydrological balance can be explored.

Relative humidity is essentially invariant in these simulations (Fig. S4); a result generic to climate models and widely considered to be an emergent property of the climate system31. Model derived soil moisture distributions are only slightly sensitive to warming, and in general in both simulations indicate soil moistures are > 60% (Fig. S5) at mid- and high latitudes. In other words, the model—in agreement with the basic physical arguments made above—place high latitudes in the net P dominated regime in which nearly saturated conditions prevail.

The model shows a mild increase in net E over the subtropical oceans, and a mild increase in net P at high latitudes (Fig. S3) with global warming, which reflects the enhanced hydrological cycle of these simulations20. As noted above, drying or moistening of a given region without the appropriate large-scale context does not provide information about the hydrological cycle (as defined here). Therefore, in order to compare model predictions with proxy interpretations, we focus on regions where we expect net P to dominate (higher latitudes) separate from region where we expect net E (low- to mid-latitudes).

Palaeocene high latitude records: The Palaeocene is commonly considered to be one of the wettest intervals in Earth history. Some of the most extensive and thickest coal deposits of the Cenozoic22,32, in western North America as well as Alaska and Northern Canada33 ( for a global map see Ziegler et al., 200422 or were deposited in the Palaeocene. Coals form from peats, which have a narrow tolerance for unsaturated conditions34 and are considered excellent indicators of conditions in which P dominated over E, leading to nearly saturated conditions35. Crucially, many of these coals, such as the “Fort Union” and “Ravenscrag” coals distributed across a large area of the northern Rockies and Northern Great Plains, are characterized by low-sulfur values. As such, they are especially indicative of year-round fully saturated conditions during their formation and formed from raised mires34. Such coals are widespread and also occur at other high-latitude locations, such as the late Palaeocene Sagavanirktok Formation in Alaska36,37.

In addition to the widespread record of perennially saturated conditions derived from the presence of coal deposits, other independent quantitative and qualitative proxies for surface paleoenvironmental and hydrological conditions are routinely interpreted as showing that the late Palaeocene was variously humid, mesic, and subtropical-to-warm temperate38-42. High-latitude vegetation during the Palaeocene was polar deciduous (mixed coniferous and angiosperm) forests. The climate regime is widely referred to as “mesothermal, humid”43. In regions where reasonably complete records exist, such as in the Canadian High Arctic (Ellesmere Island and environs), coastal regions were swamps and wetlands, conditions further inland were floodplains and braided stream plains continuing into inferred upland forests43,44 Conditions such as these are widely reflected throughout high-latitude paleoclimate records41. Other extratropical regions were characterized by evergreen broad-leaved forests with a “mixed” subtropical character41. Records of massive bursts of highly weathered clays at high latitudes45-47 can also be considered evidence of major increases in high-latitude hydrological cycling during this interval, although the exact timing and interpretation of these records is not always exact48,49.

Low- to mid-latitude records: High resolution records for the Palaeocene50,51 provide paleobotanical estimates that late Palaeocene (Clarkforkian) mean annual precipitation in Wyoming was about 130 cm/year (characterized as ‘humid’ subtropical rainforest). Within this region, workfocused on hydrologic and temperature variability; both temporally and spatially52-55. An early Palaeocene ‘rain forest’ was recently discovered in Colorado55. Paleoprecipitation estimates for this site are roughly double the value for contemporaneous sites located nearby—a result that may be due to orographic influence56. This distribution is consistent with our model results, but more importantly, it highlights the potential for strong regional variability in hydrological fields in this region. Thus, sites in western North America are difficult to interpret in that regard because the inferred water balance of the region is likely to be strongly dependent on topography. With that caveat in mind, we note that most of Wyoming is a region of strong net P in these simulations, immediately adjacent to regions of net E. This pattern of variability over small scales is grossly consistent with previously noted apparent moisture gradients in the region53,54. Model-predicted precipitation is ~150 cm/year in Wyoming, increasing slightly with warming (~10 cm/yr). These values compare favorably with proxy estimates51.