Integrating plant wax abundance and isotopes for paleo-vegetation and paleoclimate reconstructions: a multi-source mixing model using a Bayesian framework

Yang, Deming; Bowen, Gabriel J.

doi:https://doi.org/10.5194/cp-18-2181-2022

Articles | Volume 18, issue 10

https://doi.org/10.5194/cp-18-2181-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/cp-18-2181-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 18, issue 10

Research article

|

30 Sep 2022

Research article |

| 30 Sep 2022

Integrating plant wax abundance and isotopes for paleo-vegetation and paleoclimate reconstructions: a multi-source mixing model using a Bayesian framework

Deming Yang and Gabriel J. Bowen

Download

Final revised paper (published on 30 Sep 2022)
Supplement to the final revised paper
Preprint (discussion started on 04 Mar 2022)
Supplement to the preprint

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2022-23', Anonymous Referee #1, 20 Apr 2022

In this paper, Yang and Bowen develop a Bayesian framwork to help interpret sedimentary n-alkane distributions. The authors outline the Bayesian framework and apply this in two case studies. The authors also outline areas for future development. I have little expertise in Bayesian statistics and unable to comment on the model itself. I suspect the paper will be of interest to organic geochemists. However, the paleoclimate implications are not clear. In its current form, this would be more suitable for a specialised journal (e.g,. Organic Geochemistry).

I have three major suggestions:

1) To paper would benefit from a more thorough comparison to existing techniques (e.g., linear mixing-model approaches of Gao et al., 2011). You state that your results appear to provide alternative interpretations to the same n-alkane records – please elaborate!

2) It is essential to validate this approach in a sediment core with independent vegetation reconstructions. I would take a look at the African records published by Sarah Feakins (e.g., Feakins 2013 P3) - these include n-alkane chain length distributions, n-alkane carbon isotopes and the % of shrub, graminoids and tree pollen. This seems an ideal site to test your approach. However, I am sure there are dozens of other suitable sites.

3) The authors state that their approach could be used to assess the interpretation of associated proxies such as n-alkane δ2H. This would be a great tool for organic geochemists and paleoclimatologists. However, the authors did not explore this any further. The authors should demonstrate how their approach can refine the interpretation of leaf wax del2H records - if they can, this paper will be far more valuable to the paleoclimate community.

Citation: https://doi.org/10.5194/egusphere-2022-23-RC1
- AC1: 'Reply on RC1', Deming Yang, 31 May 2022
  
  We thank the reviewer for taking the time to review our manuscript, and the constructive suggestions. The reviewer suggested that we should elaborate on three aspects of the manuscript. Below is our response to the reviewer’s suggestions point by point.
  1) To paper would benefit from a more thorough comparison to existing techniques (e.g., linear mixing-model approaches of Gao et al., 2011). You state that your results appear to provide alternative interpretations to the same n-alkane records – please elaborate!
  In our case studies (CS1 and CS2), we did provide alternative interpretations to the published data. In CS1, we compared the conventional interpretation of aquatic plant input based on the P_aq index and our interpretation based on δ¹³C and chain length distribution of three n-alkane chains. In CS2, we compared the interpretation of vegetation composition based on δ¹³C of one n-alkane chain, and our interpretation based on δ¹³C and chain length distribution of three n-alkane chains. Both case studies provided comparisons with existing interpretation techniques and we highlight where the new method’s results are similar to those methods as well as where they provide alternative or more nuanced interpretations.
  We can not provide a direct comparison with the approach by Gao et al. (2011), specifically, based on the case studies we currently include, because these datasets do not contain all the information required by Gao’s method. As discussed in the next response, we do plan to add additional comparisons to established methods for vegetation reconstruction.
  2) It is essential to validate this approach in a sediment core with independent vegetation reconstructions. I would take a look at the African records published by Sarah Feakins (e.g., Feakins 2013 P3) - these include n-alkane chain length distributions, n-alkane carbon isotopes and the % of shrub, graminoids and tree pollen. This seems an ideal site to test your approach. However, I am sure there are dozens of other suitable sites.
  We agree that the approach would be much more convincing if the results are consistent with an independent vegetation reconstruction, e.g., by pollen analysis. We have done a literature search and identified data that will allow us to complete such a comparison. We are planning to include such a comparison in the supplementary document. We would like to note that the comparison won’t be a true validation, because the two approaches are associated with different potential biases such as pollen/n-alkane production and transportation. For this reason, some differences between the approaches are to be expected.
  3) The authors state that their approach could be used to assess the interpretation of associated proxies such as n-alkane δ²H. This would be a great tool for organic geochemists and paleoclimatologists. However, the authors did not explore this any further. The authors should demonstrate - if they can, this paper will be far more valuable to the paleoclimate community.
  We agree with the reviewer that adding information on how the approach can help to interpret associated n-alkane δ²H will add tremendous value to the significance of the framework. Doing so would make the manuscript more aligned with the aims and scope of the journal. We are planning to add one more case study to demonstrate how our framework can be implemented. The case study will be based on n-alkane records in a marine core off the Zambezi River Mouth (Wang et al., 2013, GCA), which has δ²H, δ¹³C, and chain length distribution data on n-C₂₇, n-C₂₉, n-C₃₁, and n-C₃₃ chains. Please note that doing so will require further development of the existing model structure, and discussion of model outputs. Please allow us some time to implement the case study in the revision.
  
  Citation: https://doi.org/10.5194/egusphere-2022-23-AC1
RC2:
'Comment on egusphere-2022-23', Anonymous Referee #2, 11 Jul 2022

In this manuscript, the authors use a modelling approach, based on Bayesian statistics under inclusion of chain-length distribution and carbon isotopes, to assess organic matter sources to sedimentary n-alkanes. Because of mixing processes, i.e. varying contribution of different sources to different chain-lengths, such novel approaches are needed to improve interpretation of proxy data in sediment cores (which so far mainly rely on relatively simplistic approaches).

This pilot study is well conducted and the first results are interesting and appear promising for future applications. I´m just wondering if Climate of the Past is the right journal for this work, which would probably better fit into GCA or OG. Especially under the terms, that a real exemplary application to a full paleorecord is missing (but would probably make this study too extensive).

In conclusion, this will be useful to paleoclimatologists which use organic geochemical methods, and I recommend publication.

Specific comments:

l27: I would say mid-chains are of likewise importance (just consider the many studies using Paq), i.e. I´d rather span the range to “C23 – C35”.

l59 and other: general comment: I think references should be sorted by year, i.e. here start with Collister and end with Liu and An. (relevant for whole text)

Figure 1: even they are explained elsewhere, abbreviations (FLMC, etc) should be explained here in the Figure caption

Table 1: so high Paq means high influence of aquatic sources onto C27 (visible at somewhat higher d13C in this sample). Even though the focus is on C27, C29, C31, would it make sense to report data also for C23, C25?

l117: “we expect n-alkane δ 13C to follow a group-specific distribution pattern”: I´m wondering how well this works for aquatic plants, because those have shown to be quite variable in their d13C values, even within similar species. This is also visible on the right panels in Figure 2 (pretty broad Gaussian blue curve overlay).

l245: micro or macroalgae? Macroalgae (e.g. Charophytes) further complicate the issue because they mainly produce mid-chains. Or are Chara sp. here included to macrophytes (because they are listed in the table EA-2 at github)? This needs clarification.

Table 3: is “MAP” the official abbreviation for Maximum A Posteriori probability estimates, or could an alternative shortening be used? Just because it´s easily confused with mean annual precipitation.

l313 and 313: what precisely is a “trade-off correlation”?

Supplement:

l12 and ff: it is unclear what EA-2, 3, 4 are referring to. Those tables appear in the github data sources. If this is the case, those should be referenced here in the supplement.

l16: two time “in”

Supplement tables: some “n” in n-alkanes are not italic

Citation: https://doi.org/10.5194/egusphere-2022-23-RC2
- AC2:
  'Reply on RC2', Deming Yang, 13 Jul 2022
  
  We appreciate the reviewer’s recognition of the novelty of the model approach, and the recommendation for publication. We also thank the reviewer for the detailed comments and constructive suggestions.
  The main issue raised by the reviewer is whether the manuscript is a good fit for Climate of the Past. The reviewer also mentioned that a case study analyzing a paleorecord will help to address the issue. We would like to note that reviewer 1 raised a similar issue.
  In our response to reviewer 1, we agreed to include an additional case study to demonstrate the model’s utility in analyzing a paleorecord, with n-alkane δ²H as an additional model component. We agree that adding this case study will make the manuscript of interest to both organic geochemists and paleoclimatologists and make it a much better fit for the journal. We are finalizing the case study right now.
  The specific comments by reviewer 2 seem to have raised some minor issues. We appreciate the reviewer thoroughness in the review, and we will do our best to address them in the revision.
  
  Citation: https://doi.org/10.5194/egusphere-2022-23-AC2
  - RC3: 'Reply on AC2', Anonymous Referee #2, 14 Jul 2022
    
    Dear authors,
    thanks for your responses. I´m looking forward to the revised version and to see another application to a paleorecord.
    Among the "minor issues" I´m specifically interested how Characeae were treated, which appear in the underlying datatables on GitHub. Did those macroalgae actually go into the "algae"-cluster, or were they treated as macrophytes (which they rather resemble in terms of their n-alkane patterns)?
    
    I have one other question:
    This whole approach might appear complicated to researchers which are less into modelling and statistics. I.e. who prefer -for example- to use simple compound ratios as proxies. Is there anything which can be done to "lower the hurdle", i.e. to make the approach more accessible and easy to apply to a broad range of researchers?
    This aspect includes, how gaussian endmember distributions should be treated and calculated for local applications. Does each application require an own calculation of prior distributions (according to the authors who use regional plant data for their two examples); or should it run towards using a global average (e.g. for C3 and C4 plants).
    The authors discuss this a bit in section 4.3.1 (e.g. mention the benefits that would come from a global plant lipid database), but it would be interesting to hear a bit more how this should be handled.
    
    Citation: https://doi.org/10.5194/egusphere-2022-23-RC3
    
    AC3: 'Reply on RC3', Deming Yang, 14 Jul 2022
    
    We thank the reviewer for the follow-up questions and thoughts.
    
    The reviewer asked about the treatment of Characeae (genus Chara) in our data sheet shared on GitHub. The genus Chara is treated as algae while we did not specify whether it is micro or macro algae. This treatment is following Aichner et al., 2010 and Liu & Liu, 2016. When we were putting the data together, we observed that the genus Chara display much lower δ¹³C values than macrophytes as reported by Aichner et al., 2010. This further justifies the reason why Chara is not grouped with macrophytes, which display higher and more variable δ¹³C values according to Aichner et al., 2010. We can certainly elaborate on this decision in the revised manuscript.
    
    The reviewer also asked about how to make this approach more accessible to a broad range of researchers, as the approach may seem rather complicated. We would like to add to this question by emphasizing the main benefits of using such a Bayesian approach, which are 1) getting the most out of compound-specific isotope analysis by offering mixing solutions of multiple sources, and 2) straightforward handling of uncertainty in the analysis. It is intuitive that researchers would only choose to use this approach if the benefits outweigh the “costs”. We would like to start with the notion that there are potentially huge benefits in using this approach, some of which are outlined in section 4.2. Case study 3 will specifically demonstrate the benefits of handling uncertainty and add to what is already in section 4.2. We really hope that this publication can increase the awareness on the benefits that this approach may bring to the community.
    
    In terms of lowering the hurdle (cost), there are several things that can be done. One is to develop more user-friendly versions of the code into perhaps an R package and distribute it with a detailed tutorial. Our longer-term plan is to do this, combining the work presented here with other proxy system models under development and providing a common interface and framework for proxy data interpretation. Such a development is premature at this point, however, and one of our short-term goals is to receive community feedback on the approach and how it is/may be useful to the research community through this and other publications. Another way to lower the ‘bar to entry’ is to organize workshops online or at conferences for the technical details of this approach if there is enough interest. Again, community feedback is key. We are also open to inquiries and potential collaborations if necessary.
    
    The reviewer also asked about the handling of prior distributions with examples of whether to use regional compilations or global averages. The model itself is flexible enough to accommodate either choice, so it ultimately depends on the user and the reasons to justify it. We chose to use regional compilations in case studies 1 and 2 because we think that this is more consistent with what we know about the production, transport and mixing processes of plant wax lipids in lake sediments. If using global averages can be justified in certain situations, we see no reason why they cannot be used. Related to this and the reviewer’s last comment, we think that having a centralized lipid database is important because it can make this model approach much easier to use from a user’s perspective, whether the need is for a regional compilation as demonstrated, or a global one.
    
    Citation: https://doi.org/10.5194/egusphere-2022-23-AC3

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Publish subject to minor revisions (review by editor) (02 Aug 2022) by Julie Loisel

AR by Deming Yang on behalf of the Authors (08 Sep 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (13 Sep 2022) by Julie Loisel

AR by Deming Yang on behalf of the Authors (13 Sep 2022) Manuscript

Short summary

Plant wax lipid ratios and their isotopes are used in vegetation and paleoclimate reconstructions. While studies often use either type of data, both can inform the mixing pattern of source plants. We developed a statistic model that evaluates ratios and isotopes together. Through case studies, we showed that the approach allows more detailed interpretations of vegetation and paleoclimate than traditional methods. This evolving framework can include more geochemical tracers in the future.