Alkenone temperature and carbon isotope records: Temporal resolution, offsets, and regionality



[1] Over the last decade the number of paleoceanographic studies presenting and discussing alkenone-derived sea surface temperature (SST) records and carbon isotope ratios of alkenones (δ13Calkenones) has increased rapidly. This short review paper summarizes the state of knowledge about SST variability determined by alkenone paleothermometry in sediment trap studies and investigations of marine sediments for timescales of single to millions of years. In particular from the evaluation of Quaternary records, characteristic temperature pattern have emerged that mismatch with SST curves estimated by other methods, e.g., foraminiferal stable oxygen isotopes and transfer functions. In future studies the spatial distribution of such variance patterns typically observed in alkenone SST curves should be checked in order to better understand their meaning for ocean circulation changes and a possible relationship to global climate change. On the other hand, certain physical and biological processes have the potential to put the reliability of alkenone SST curves into question. Accordingly, examples for possible effects of bioturbational mixing, lateral transport of remote alkenones, diagnetic alteration and degradation of alkenone composition, biogeographic and temporal diversity changes in alkenone-producing algae, as well as the influence of changes in salinity and nutrient conditions on the alkenone paleotemperature signal, are briefly mentioned, highlighting the need for more detailed future studies on these issues. Finally, a short summary on the status of the use of the δ13Calkenones signal as a proxy for past variations in the carbon dioxide pressure (PCO2) in ocean surface waters and related problems is given.

1. Introduction

[2] Records of Quaternary sea surface temperatures (SSTs) estimated from ketone unsaturation ratios and of δ13Calkenones are available for most of the major ocean basins. A few extend into the late Neogene. The application of the alkenone U37K′ or U37K index [Brassell et al., 1986; Prahl and Wakeham, 1987; Rosell-Melé et al., 1995] for estimation of paleotemperatures from sedimentary molecular compounds of marine phytoplankton is an increasingly applied tool for modern concepts in paleoceanography and paleoclimatology. However, its application has mainly concentrated on various timescales for the Holocene and the late Pleistocene so far. Alkenone-based estimates of variations in SSTs do not always agree with those derived from transfer functions based on microfossil assemblages (see discussion by Bard [2001]. Accordingly, questions have been raised about potential biases of the alkenone method related to sedimentation processes through the water column and at the seafloor, which may alter the alkenone temperature signal independently from ecological and environmental factors acting in surface waters [see Volkman, 2000]. The first aim of this contribution is to present an overview on alkenone SST and δ13Calkenone records available from the literature and from own work for the different historical and geological timescales of interest, e.g., annual to decadal records and records covering several thousands to millions of years. Second, the SST records are evaluated for any patterns of variability and offsets inherent in the alkenone SST curves that may be characteristic either for certain regions in the ocean, for certain time periods, or for certain types of sediments, e.g., high and low sedimentation rates, laminated or bioturbated, anoxic or oxic conditions, and lateral suspension transport. In constrast to the large set of alkenone SST curves available, only a few Neogene δ13Calkenones records have been published to date. Therefore this contribution focuses on a brief summary describing what is available and what the concerns are when using the δ13Calkenones signal for calculating past levels of dissolved carbon dioxide for the surface ocean.

[3] A more expanded version of the content of this paper was presented as a status report for discussion in the Working Group 5 “Paleoceanographic Interpretation, Temporal resolution, Offsets and Regionality” of the National Science Foundation (NSF) sponsored workshop on “Alkenone-based Paleoceanographic Indicators” held at Woods Hole Oceanographic Institution during October 1999. A comprehensive report on the outcome of the discussion will be presented elsewhere [Sachs et al., 2000]. The working group report will concentrate in much more detail on the importance of the physical and biological processes in the surface ocean and at the seafloor that could have had a biasing effect on the alkenone distribution and composition in the sediments, while this contribution will mainly formulate the questions arising from evaluation of typically observed Quaternary and Neogene SST and δ13Calkenones patterns.

2. Status and Discussion

2.1. Holocene to Late Neogene Alkenone SST Variability and Spatial Patterns

[4] For modern times and the latest Holocene, SST records from sediment trap series [Prahl et al., 1993; Treppke et al., 1996; Ternois et al., 1997; Conte et al., 1998; Müller and Fischer, 2000], as well as from laminated sediments and from sites with very high sedimentation, reveal the variability of surface ocean temperatures with a resolution of years to decades [McCaffrey et al., 1990; Kennedy and Brassel, 1992; Herbert et al., 1998]. Such records indicate SSTs that strongly vary seasonally (e.g., a range of 7°C off Namibia, as shown in Figure 1), and annually. At the respective sites these alkenone estimates match the hydrographic measurements of intra- and interannual variations reasonably well [e.g., Müller and Fischer, 2000]. The great potential of these records is the possibility to compare the alkenone SST estimates from sinking particles and the upper sediment column with measurements of ambient temperatures from oceanographic surveys and historical records (Figure 2).

Figure 1.

Annual SST variations (solid line) as estimated from alkenones in sediment-trap samples off Namibia, Walvis Ridge, compared to variations in meridional wind stress component (dotted line) and flux rates of diatom valves. Coldest temperatures indicate intensified upwelling that co-occurs with intervals of high rates of productivity and export fluxes of diatoms, both lagging wind stress maxima by about four weeks due to the period needed for sinking to water depths of 600 m [from Treppke et al., 1996].

Figure 2.

Santa Barbara Basin alkenone-derived SST record compared to the historical temperature record from Scripps Pier (a) and variations in alkenone content (b) [from Herbert et al., 1998].

[5] The variability of SSTs throughout the Holocene can be viewed with fairly high temporal resolution at sites on continental margins or in marginal seas. For instance at sites off Peru and in the Santa Barbara Basin, the range of Holocene temperature variation reaches 2°–3°C on the basis of annual mean values [McCaffrey et al., 1990; Kennedy and Brassel, 1992; Herbert et al., 1998]. However, most of the records with temporal resolution of a few centuries only show SST variations of ∼0.5° to 1°C within the Holocene [Sikes and Keigwin, 1994; Chapman et al., 1996; Bard et al., 1997; Rosell-Melé, 1998; Cacho et al., 2000; Pelejero et al., 1999].

[6] Many contributions using alkenone paleothermometry have focused on the SST decrease during the Last Glacial Maximum (LGM). The results are summarized in the SST map produced by the TEMPUS project (http:/, which summarizes differences between temperatures in the modern surface ocean and the alkenone SST values estimated for the LGM [Rosell-Melé et al., 1998]. The most remarkable feature in this map is the general difference in the degree of cooling between the hemispheres. When compared to annual mean modern atlas values, alkenone SSTs indicate a strong LGM cooling of up to 10°C in the Northern Hemisphere, but only 3°–5° in most areas of the Southern Hemisphere [Schneider et al., 1996; Bard et al., 1997; Kirst et al., 1999] including the subantarctic ocean [e.g., Ikehara et al., 1997]. The main problem with respect to the LGM temperature reconstructions is their reliability in areas with very low alkenone content where often the sedimentation rates are also very low, e.g., the western Pacific [Ohkouchi et al., 1994]. For such regions it is not yet clear whether the alkenone-based technique is fully satisfactory.

[7] More recently, reconstructions of SST variations on the timescale of centuries have become available for the late Pleistocene [Zhao et al., 1995; Bard et al., 1997; Kirst et al., 1999; Cacho et al., 2000; Pelejero et al., 1999; Rühlemann et al., 1999; Sachs and Lehman, 1999]. These are of particular interest because of the magnitudes of temperature changes indicated and because of their temporal relationship to marine oxygen isotope and ice core records showing similar variations in their signal (Figure 3).

Figure 3.

Alkenone-derived SST (a) variations in the Alboran Sea compared to (b) variations in the δ18O composition of the planktonic foraminifera G. bulloides and to changes in the abundance of polar foraminifera species N. pachyderma (sin.). In particular, the alkenone temperature and the N. pachderma record indicate pronounced cooling events (∼2° to 3°C decrease) at submillenial timescales for the Mediterranean Sea probably coinciding to cold spells recorded in Greenland ice cores (d) [from Cacho et al., 2000].

[8] Numerous reports have presented late Quaternary SST records with a temporal resolution of a few kiloyears. Within these records, patterns have emerged that are distinct from those resulting from isotopic records or from transfer functions based on assemblages of microfossils. These are (1) occurrence of coldest SST not at the end of glaciations but 10–15 kyr earlier; (2) relatively warm tropical and subtropical SSTs in marine isotope stage 6 [e.g., Eglinton et al., 1992; Rostek et al., 1993; Emeis et al., 1995a; Schneider et al., 1995, 1996; Villanueva et al., 1998a; Budziak, 2000]; (3) a general warming trend over the last 500 kyr, with very cold glacial SST at the beginning of the Brunhes Chron and by ∼1°C warmer at the end; and (4) lacking evidence for relative warmth during marine isotope stage 11, in the middle of the Brunhes Chron [Brassell et al., 1986; Emeis et al., 1995b]. The spatial patterns of these distinct features in the alkenone SST records within the world ocean remain to be determined. They have the potential to reveal unknown patterns of past ocean circulation and thus to suggest mechanisms of global climate change. A first attempt to explain the first two aspects of a relatively warm glacial stage 6 and midglacial cooling in the tropics (Figure 4) as a direct response to climate forcing in low latitudes is provided in the work of Schneider et al. [2000].

Figure 4.

Alkenone-derived SST record (line with black dots) from the Angola Basin showing the temperature pattern typical for the tropical Atlantic ocean and also found at sites in the Indian and Pacific Ocean. Superimposed the eccentricity component of variation in orbital configuration is shown (black line). Note that very cold temperatures correlate with minima in eccentricity and not with full glacial conditions as commonly indicated by oxygen isotope records [from Schneider et al., 2000].

[9] For late Neogene times, only very few continuous SST records based on alkenone paleothermometry exist from Ocean Drilling Program (ODP) sites in the North Pacific [Haug, 1996], the Canary Current [Herbert and Schuffert, 1998], and the Benguela Current [Marlow et al., 2000]. Interestingly, both records from the eastern boundary currents in the Atlantic suggest ∼6°–8°C higher surface water temperatures during the late Miocene/early Pliocene compared to Holocene temperature values. However, the reliability of the alkenone indices as SST indicators prior the Late Quaternary is still a matter of debate. To compare the result from alkenone paleothermometry with those from other methods (e.g., foraminifera oxygen isotopes or Mg/Ca ratios) may help to evaluate the potential of the alkenone unsaturation ratios for the Neogene temperature reconstructions.

2.2. Potential Effects of Physical and Biological Processes on the Alkenone SST Signal

[10] First, as with all other temperature proxies, amplitudes of signals and their significance relative to noise levels must be evaluated on different timescales. Relationships to bioturbation and to the temporal resolution of sampling are of particular interest. Often, alkenone-based SST records are, in general, less noisy than those based on other methods and the amplitude of temperature changes is smaller than indicated by other SST proxy records. An extreme example is shown in Figure 5 for the western Pacific, where the magnitude of the LGM to Holocene SST increase is very small [Ohkouchi et al., 1994]. It is possible that the original LGM to Holocene SST contrast has been smoothed out owing to bioturbational mixing and very low sedimentation rates of ∼12 mm/kyr. The main question here is whether bioturbation has a stronger “smoothing effect” [e.g., Gong and Hollander, 1999] on alkenone SST variability compared to methods based on coarser grained microfossils (E. Bard, Paleoceanographic implications of the difference in deep-sea sediment mixing between large and fine particles, submitted to Paleoceanography, 2000b). Moreover, in areas with very low alkenone content or with strong bottom current activity the sediments can contain a relatively high amount of old reworked alkenones (A. Rosell-Melé, personal communication, 2000) or of suspended material laterally transported to the site of investigation [Benthien and Müller, 2000]. The degree to which such processes can corrupt local alkenone signals is not well known.

Figure 5.

Alkenone SST record from Western Tropical Pacific [from Ohkouchi et al., 1994].

[11] Second, on Milankovitch and sub-Milankovitch timescales, distinct SST pattern occur in tropical and high latitudes of the Southern Hemisphere that were shortly described above (patterns 1 and 2, see also Figure 4) and should be discussed for their regional or basin-wide significance. Despite some concerns on the reliability of alkenone SST estimates it turns out that the pattern typical for the tropics can be found in all three oceans [e.g., Schneider et al., 2000]. The question here is whether features such as the late Quaternary patterns (patterns 1–4, above) result in some way from ecological or diagenetic biases or from some other artefact. Comparisons of regional aspects of these features with biogeographic provinces should be helpful. Comparisons to estimated paleochemical variations (availability of nutrients, oxygenation of sediments) may also be informative.

[12] Third, analyses of algal cultures grown at different temperatures show that different strains of Emiliania huxleyi and different species of haptophytes can yield different relationships between U37K′ and temperature (see Müller et al. [1998] and Volkman [2000] for discussion). Also, an increase in the abundance of E. huxleyi concommitant with a decrease of Gephyrocapsa spp. over the last 60 kyr resulted in generally low amounts of C37 alkenones in Atlantic and Indian Ocean sediments [Müller et al., 1997; Rostek et al., 1997]. The latter (see Figure 6) has implications for paleoproductivity reconstructions based on total amounts of alkenones in marine sediments [e.g., Rostek et al., 1997; Villanueva et al., 1998b; Budziak et al., 2000]. If differences in alkenone production and accumulation rates depend on haptophyte taxa changes over time, then variations of alkenone accumulation rates can only be used with caution as a proxy for past changes in total haptophyte productivity or marine productivity in general. Accordingly, paleo-SST records should be tested for signals which may not be driven by ocean temperature changes but by significant changes in the composition of haptophytes/coccolithophorids over time or by insufficient amounts of C37 alkenones produced. Such studies should indicate the extent to which species- and concentration-related factors can influence values of U37K′. They are pertinent also to considerations of the applicability of modern relationships between alkenone unsaturation ratios and temperature to alkenones found in ancient sediments.

Figure 6.

SST, U37K′ index, and alkenone content compared to relative abundance changes in two coccolith species dominating the sediments at the Walvis Ridge, SE Atlantic. Note the decrease in overall alkenone content when G. huxleyi contents increase toward the Recent [from Müller et al., 1997]. Copyright 2000 with permission from Elsevier Science.

[13] Fourth, we must consider whether geographic variations might reveal that unsaturation ratios can be affected by changes in salinity or in concentrations of nutrients. Paleo-SST records can be examined for correlations with past changes in salinity, concentrations of nutrients, and primary productivity [Emeis et al., 1995a; Summerhayes et al., 1995; Kirst et al., 1999]. Rosell-Melé [1998] has proposed that records from the North Atlantic (Figure 7) indicate that U37K can depend on salinity. Up to now, however, no evidence for such effects has been recognized near the mouth of large rivers entering the tropical ocean. For past changes in nutrient levels (and thus perhaps for changes in rates of growth), there is no definitive evidence that alkenone unsaturation ratios are affected in the sense implied by culture experiments: low nutrient contents induce higher U37K′ [Epstein et al., 1998]. With respect to this problem, comparisons of paleo U37K′ records with proxy records revealing past nutrient conditions or paleoproduction at the same site [e.g., Lyle et al., 1992; Holmes et al., 1997] should be considered in more detail (Figure 8).

Figure 7.

Anomalous warming in the Younger Dryas period according to the alkenone U37K index (b) in core HM 79-4/6 off southeastern Norway is paralleled by very high amounts of C37:4 (a) which may be related to lowered salinities at that period. (c) For comparison, SSŤs estimated by diatom transfer function (DTF) for the summer season are also shown. Note that the warm temperature anomaly during the Younger Dryas suggested by the alkenone method does not appear in the DTF appoach [from Rosell-Melé, 1998].

Figure 8.

Alkenone SST, δ15N, and paleoproductivity records from the Angola Basin, SE Atlantic, for (a) Congo fan and (b) Angola margin sediments [from Holmes et al., 1997]. Comparison of variations in nitrogen isotope ratios of bulk organic matter with changes in alkenone temperatures may allow the test whether ketone unsaturation rations depend on nutrient conditions as suggested by tank experiments using haptophyte cultures.

[14] Finally, possible effects of diagenesis and biodegradation on unsaturation ratios should be considered. This problem is twofold. On the one hand, effects of varying levels of O2, and thus of varying diagenetic pathways, should be examined. On the other hand, the possibility that differential degradation might cause variations in U37K′ that become significant over longer periods of time should also be considered. For the first of these the studies of Prahl et al. [1989] and Gong and Hollander [1999] are significant. Similarly, records of SST through Mediterranean sapropel events provide a basis for discussion [Emeis et al., 1998; Doose et al., 1999]. To rule out longer-term diagenetic effects, the several million-year long records from Haug [1996], Herbert and Schuffert [1998], and Marlow et al. [2000] provide the opportunity to try to validate alkenone SST estimates for the Neogene by other methods.

2.3. Paleo- pCO2 Reconstruction Using δ13C(alkenones)-

[15] Records of the carbon isotopic composition of alkenones (δ13Calkenones) and their interpretation are much rarer than those of alkenone unsaturation ratios. For the Late Quaternary and the Miocene this topic has been addressed by several authors in order to reconstruct partial pressure of carbon dioxide for the ancient ocean. For the late Quaternary, only a few records are available for the Gulf of Mexico [Jasper and Hayes, 1990], for the eastern Pacific [Jasper et al., 1994], and for the eastern Atlantic [Andersen et al., 1999]. The use of these records for reconstruction of pCO2 raises several issues which deserve attention.

[16] The isotopic compositions of alkenones relative to dissolved CO2 are affected not only by pCO2 but also by rate of growth and by the size and shape (ratio of surface area to volume) of the algal cells [Popp et al., 1999]. Unfortunately, there are at present only semiquantitative, or probably qualitative, independent geochemical and biological paleoindicators for these environmental variables. At the moment, it is very difficult to see how past variations in δ13Calkenones can be securely interpreted in terms of variations of (only) pCO2 or growth rate (e.g., A. Benthien et al. The carbon isotopic composition of C37:2 alkenones in core top sediments of the South Atlantic: Effects of CO2 and nutrient concentrations, Global Biogeochemical Cycles, 2000). However, based on results to date, it is worth mentioning that most of the variations in the values of surface water pCO2 estimated for the Neogene are introduced by the paleotemperature estimates used and not so much by the observed variability in the δ13Calkenones. [e.g., Andersen et al., 1999; Pagani et al., 1999a, 1999b]. Therefore it is essential to consider not only the variations of δ13Calkenones back in time (Figure 9) but also the total range of δ13C changes relative to amplitude of the SST changes when calculating past pCO2 values for the surface ocean. So the question comes back to the most reliable method for SST estimates. To obtain a view of the importance of any carbon-isotopic paleobarometer, it is necessary to determine first the relative importances of SST and concentrations of dissolved CO2 in controlling atmospheric pCO2. Additionally, studies of boron isotopic abundances and of oceanic pH (and thus of the speciation of dissolved inorganic carbon [Sanyal and Bijma, 1999]) may help to understand the significance of observed variations in δ13C(alkenones). Further points of discussion for the paleo-pCO2 reconstructions based on δ13C values should be methodological problems, the lack of records with higher temporal resolution, and other marine biomarkers that may have the potential to reconstruct the concentration of dissolved carbon dioxide [CO2aq]. For discussion of Tertiary pCO2 reconstructions based on δ13C of specific markers we should rely on the reconstructions of Freeman and Hayes [1992] and Pagani et al. [1999a, 1999b].

Figure 9.

Miocene changes in atmospheric carbon dioxide pressure (pCO2) as estimated from δ13C37:2 values in Ocean Drilling Program cores from the West Pacific (site 588), the western Arabian Sea (site 730), and the subtropical North Atlantic (site 608). Solid and stippled lines relate to different equations used for the relationship between isotopic fractionation and dissolved carbon dioxide concentration [in surface waters CO2aq]. For details, see Pagani et al. [1999a]. Note that atmospheric pCO2 is presented, although empirical relationships and surface ocean temperature estimates used by Pagani et al. [1999a] would only allow estimation of surface-ocean carbon dioxide pressure (PCO2) levels.

3. Concluding Remarks

[17] It is difficult to give conclusion at the end of such a status report and after a workshop discussion that raised a suite of open questions to be addressed in future studies. Recommendations for such studies will be lined out in the working group report [Sachs et al., 2000]. Without doubts the alkenones as a new molecular tool have offered a tremendous potential for new paleoceanographic studies on past temperature and chemical conditions in the surface ocean. Determination of alkenone composition and their carbon isotopic signature already provided records that definitively gave new insights and better understanding of climate change at the various historical and geological timescales the community is interested in. For this reason any new record from the water column and from sediments will enable to further test the applicability of alkenones as palaeoceanographic indicator, particularly, when studied in context with other paleoenvironmental parameters, and will provide new times series for past chemical and physical conditions in the surface ocean. This should be important enough to proceed with the application of alkenones as a paleoceanographic tool, despite all criticisms against the use of the sedimentary alkenone signal to determine past temperatures, CO2 levels, and/or nutrient conditions.


[18] I would like to thank T. Eglinton, J. Hayes, M. Conte, and G. Eglinton for their invitation to attend the NSF Workshop “Alkenone-based Paleoceanographic Indicators” at Woods Hole, October 1999. This contribution significantly benefitted from reviews by F. Rostek, M. Ikehara, and J. Hayes and emerged from paleoceanographic studies carried out together with P. Müller at the Fachbereich Geowissenschaften, Bremen University, supported by grants from the Deutsche Forschungsgemeinschaft Sonderforschungsbereich (SFB) 261 and the European Community project ENV4-CT97-0564. This is SFB 261 contribution 314.