Lasers, biomarkers and the Sun

The earliest work on Holocene palaeoecology focused on megafossils such as Pinus stumps. Then macrofossils such as hazel shells were used to reconstruct species distributions and climate. Then pollen analysis became important, complementing rather than supplanting the larger fossils. The end of this progress towards smaller and smaller proxies is found with biomarkers, chemical tracers of species presence and their environment.

Not only can the proxies used become smaller, the size of the sample required for the proxy can become smaller. This is hugely advantageous, allowing high resolution work and reducing competition for mud from a core. Perhaps the ultimate in small samples has been achieved with the nanogram samples in a new paper in PNAS by Wörmer et al (2014).

Wörmer et al shine a laser onto their sediment which causes organic biomarkers to be released from the sediment so they can be measured by a fancy mass spectrometer. The laser makes sub-millimetre spots on the surface of a core (as small as 10 µm) allowing extremely high resolution sampling, even in systems with low sedimentations rates.

Wörmer et al analyse some biomarkers produced by Archaea. These biomarkers are used by the Archaea in their cell membranes – the ratio of the different biomarkers changes with the temperature at which the cells grew to maintain membrane fluidity. Consequently,  the ratio of the different biomarkers can be used to infer past temperature. The ratio used by Wörmer et al is similar to the TEX86 temperature proxy (there are technical reasons why TEX86 cannot be used).

Obviously, if you are going to work on sediment cores at a very high resolution, you need cores that are undisturbed. If, for example, animals have burrowed through the sediment, mixing it, the high frequency variability will have been smeared out. Animals need oxygen to live, so sediments deposited in anoxic conditions will have little or no bioturbation, preserving the high frequency variability.

To demonstrate their methodology, Wörmer et al analysed a section of core from a period in the early Holocene when the eastern Mediterranean was anoxic. This anoxia occurred because higher precipitation in the Nile catchment led to fresher surface conditions. As fresh water is less dense than salty water, this made it more difficult to mix the surface and deep waters, so not enough oxygen could be supplied to the deep water. The resulting anoxic sediments are known as sapropels. Several are known from the Mediterranean; Wörmer et al analyse the most recent with a resolution of about 4 years.

Their sea surface temperature (SST) reconstruction is surprisingly dynamic. Thirty year SST means range between 24.1 and 30.7 °C. The authors note that this variability might be partially due to factors other that changes in SST, including changes in the ecotype of Archaea present, changing seasonality of production and changing depth (and hence temperature) of the chemocline where enhanced production of the biomarkers occurs.

Part of the problem is that Archaea ecology is poorly constrained. When early Scandinavian palaeoecologists found Pinus stumps in peat bogs the interpretation was obvious – Pinus once grew here, and we know a lot about Pinus ecology. Interpretation of pollen data is more complex – different species have different production rates and disperse pollen over different distances, but these factors are well know and fairly well constrained. Biomarkers produced by little known microbes give lots of great data, but the interpretation is hard.

Wörmer et al run a spectral analysis on their reconstruction and declare that they find evidence of solar variability in the data. This is where, as you might imagine if you have read my previous posts on palaeoecological evidence of solar variability, I cease to praise the paper.

Wörmer et al figure 3 (A) Downcore profile of CCaT and seven-point running average (red line). Data points are mean of ∼15 measurements. (B) Spectral analysis for the downcore CCaT values, theoretical red noise (dashed line) and 99% false alarm level (dotted line). (C) Mean-subtracted downcore profile of CCaT overlaid with band-pass filtered signals centered at a frequency of 0.79 cycles cm−1 (blue line).

Wörmer et al figure 3 (A) Downcore profile of CCaT and seven-point running average (red line). Data points are mean of ∼15 measurements. (B) Spectral analysis for the downcore CCaT values, theoretical red noise (dashed line) and 99% false alarm level (dotted line). (C) Mean-subtracted downcore profile of CCaT overlaid with band-pass filtered signals centred at a frequency of 0.79 cycles cm−1 (blue line).

There is one statistically-significant spectral peak with a periodicity of 212 years, similar to the de Vries ~200 year cycle. There is no evidence of a ~90 year Gleissberg cycle – there is a trough in the spectrum at the relevant frequency (1.8/cm). This might be thought a little odd – why should the proxy be sensitive to solar variability at one frequency but not another? Wörmer et al obviously thought so to, for they cite a modelling study, Seidenglanz et al 2012, as showing “the potential of the ∼200-y de Vries cycle, but not of the ∼90-y Gleissberg cycle, to impact both surface and middepth water temperature in the Mediterranean Sea.”

Seidenglanz et al run a climate model forced by either 200 or 90 year periodicity in solar irradiance and identify regions of the modelled ocean that have coherent temperature variability. It is an interesting paper. However there are problems for Wörmer et al. First, however useful climate models are at a global level, they are less reliable at a regional scale. Seidenglanz et al use a low resolution model (3.6°). This is minimal for work in the Eastern Mediterranean where there are very few grid boxes. Such a low resolution model cannot be expected to fully capture the oceanographic process in the eastern Mediterranean. Second, zooming right in to the figures in Seidenglanz et al, it would appear that the coherence at the surface is only significant for the 90 year cycle, and not for either cycle at mid-depths. All told, the Eastern Mediterranean results from Seidenglanz et al do not help Wörmer et al, but the ideas they present that different periodicities of solar variability might be detected in different proxy records is.

Any spectral analysis of proxy data is dependent on the precision of the chronology. However, I will not criticise the age-depth model for core GeoB 15103–1 analysed by Wörmer et al. There isn’t one. This is why the spectrum in figure 3 has the axis label cm-1 rather than the expected yr-1. Details of how Wörmer et al construct a chronology for a core with no dates are given in the supplementary material.

The unoxidized layer of sapropel 1 at our station is about 19 cm thick according to element concentrations obtained by XRF scanning (7, 8) (Fig. S9) and thus slightly thicker than most sapropel 1 layers (e.g., ref. 9). The pore water profile of Mn clearly indicates that the postdepositional oxidation (burn down) is still ongoing. The formation of sapropel 1 in the eastern Mediterranean is considered a basinwide synchronous event (10), and the unoxidized sapropel has been described to cover an average of 3,000 y of sediment deposition (e.g., ref. 11). Assuming a continuous deposition, this translates into a linear sedimentation rate of about 6.3 cm ky-1. This value is in general agreement with a sedimentation rate of 6.6 cm ky-1 for the interval between 6.31 and 7.60 ky [23 and 32 cm below seafloor (cmbsf); corresponding to the upper section of sapropel 1] derived from the age determination by Paterne [data deposited in Pangaea database, www.pangaea.de (doi: 10.1594/PANGAEA.407609)] for sediment core MD84-641 taken at immediate proximity to our study site.

There are problems here that mean that the mean sedimentation rate of 6.3 cm ky-1 is uncertain and hence the cycle length of 200 years is uncertain.

Rossignol-Strick (1999) (ref 11) argue that the deposition of the sapropel took place over about 3000 years (from 9000 to 6000 BP). However the sapropel can be oxidised after deposition and the remaining unoxidised sapropelic sediment can represent a shorter interval. At one site in the Levantine basin the top of the sapropel is at 7650 BP.

At MD84-641, adjacent to GeoB 15103–1, the top of the sapropel is 6500 BP. That would make the sapropel 2500 years long, a mean sedimentation rate of 7.6 cm ky-1, and a cycle length of 175 years.

The reference to Paterne for MD84-641 sends us right down the rabbit hole. The Pangaea.de page for this core links to a paper by Paterne discussing cores from the Tyrrhenian Sea (the other side of Italy). I’ve not managed to find the correct reference for the chronology of MD84-641, but it might be Fontugne et al (1989) which I have not managed to find online.

The sedimentation rate between 23 cm and 32 cm in core MD84-641 is not constant. It varies between 2.3 and 11.7 cm ky-1. Some of this variance will be due to uncertainty in the dates, but I think some is real and that the sedimentation rate is not constant. There is certainly enough uncertainty about the sedimentation rate of the small section of the sapropel in GeoB 15103–1 to be cautious that the cycle length matches the de Vries cycle. (NB – the dates are probably not calibrated, but that won’t change the argument much)

Enough with the chronology. Wörmer et al estimate their spectrum with REDFIT, a method suitable for unevenly spaced samples (although because they assume a constant sedimentation rate their samples are actually evenly spaced). The significance test in REDFIT assumes that the proxy comes from an AR(1) process. If this is not true, REDFIT cannot be expected to give unbiased estimates of the significance of spectral peaks. There is a test in REDFIT of whether a proxy record comes from an AR(1) process. Wörmer et al don’t use it. To be fair, I don’t think anybody has ever used it. Yes, most if not all of the papers using REDFIT to identify solar cycles in proxy data have not tested whether the assumptions of REDFIT are met. Oops. The spectrum in Wörmer et al looks suspiciously like what one would expect if the underlying process is not an AR(1) process. I’ll write more about this in the future – a couple of manuscripts are in preparation.

One last point. Wörmer et al identify a de Vries cycle in their proxy data during the late phase of sapropel 1, sometime between 8000 and 6000 BP. This is a period when there is no significant variability at the de Vries time scale in the cosmic isotope record of solar activity by Steinhilber et al (2012). It would seem unlikely that archaeal biomarkers would record a solar cycle not detectable by cosmic isotopes.

Comparison of solar activity (blue) and δ18O from Dongge cave, China (green). both records are detrended.  (A) Time series of solar activity (TSI) and δ18O. (B) Wavelet of solar activity (TSI).  Black boundaries mark 95% significance level. (C) Wavelet coherence of solar activity (TSI) and δ18O. De Vries cycle at approximately 210 y and Eddy cycle at approximately 1,000 y are marked with horizontal, gray dashed lines. Arrows pointing to the right indicate that the records are in phase. Black boundaries mark the 95% significance level.

Comparison of solar activity (blue) and δ18O from Dongge cave, China (green). both records are detrended. (A) Time series of solar activity (TSI) and δ18O. (B) Wavelet of solar activity (TSI). Black boundaries mark 95% significance level. (C) Wavelet coherence of solar activity (TSI) and δ18O. De Vries cycle at approximately 210 y and Eddy cycle at approximately 1,000 y are marked with horizontal, gray dashed lines.
Arrows pointing to the right indicate that the records are in phase.

The analytical biogeochemistry in Wörmer et al is excellent. However, the claim of solar variability in the biomarker record is not credible. I strongly suspect that the spectral peak is an artefact of the methods used, and in any case its period is too uncertain to assign to solar variability with confidence.

You can expect this paper to be cited by the next edition of the anti-IPCC report, the NIPCC.


Fontugne et al (1989) Initiation de la stratification de la Mediterrane orientale et debit du Nil a l’Holocene, in Past and Future Evolution of Deserts, Actes Colloque Prog. Int. Corr. Geol.(IGPC) 252, Jerba, Tunisia

Wörmer et al 2014. Ultra-high-resolution paleoenvironmental records via direct laser-based analysis of lipid biomarkers in sediment core samples. PNAS

Advertisements

About richard telford

Ecologist with interests in quantitative methods and palaeoenvironments
This entry was posted in climate, Novel proxies, Peer reviewed literature, solar variability and tagged , , . Bookmark the permalink.

One Response to Lasers, biomarkers and the Sun

  1. A group of us at the Azimuth Project are analyzing current records of ENSO to see if we can pick up patterns. Here is one such thread:
    http://azimuth.mathforge.org/discussion/1504/2/symbolic-regression-machine-learning-and-enso-time-series

    The Wormer work is very interesting. The ENSO behavior might have a solar cycle component that modulates the other forcing factors. The long term paleo work can pick up these factors on a much longer time-scale.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s