Differing pre-industrial cooling trends between tree rings and lower-resolution temperature proxies

Klippel, Lara; St. George, Scott; Büntgen, Ulf; Krusic, Paul J.; Esper, Jan

doi:https://doi.org/10.5194/cp-16-729-2020

Articles | Volume 16, issue 2

https://doi.org/10.5194/cp-16-729-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/cp-16-729-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 16, issue 2

Research article

|

09 Apr 2020

Research article |

| 09 Apr 2020

Differing pre-industrial cooling trends between tree rings and lower-resolution temperature proxies

Lara Klippel, Scott St. George, Ulf Büntgen, Paul J. Krusic, and Jan Esper

Download

Final revised paper (published on 09 Apr 2020)
Supplement to the final revised paper
Preprint (discussion started on 04 Apr 2019)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Review of MS No.: cp-2019-41 Klippel et al. Research Article', Edward Cook, 25 Apr 2019
- AC1: 'Point-by-Point Response - Review I', Lara Klippel, 09 Jul 2019
RC2: 'Review to Klippel et al.', Lea Schneider, 16 May 2019
- AC2: 'Point-by-Point Response - Review II', Lara Klippel, 09 Jul 2019
RC3: 'Review of Klippel et al.', Anonymous Referee #3, 17 May 2019
- AC3: 'Point-by-Point Response - Review III', Lara Klippel, 09 Jul 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (10 Jul 2019) by Jürg Luterbacher

AR by Lara Klippel on behalf of the Authors (18 Jul 2019) Author's response Manuscript

ED: Reconsider after major revisions (19 Jul 2019) by Jürg Luterbacher

ED: Referee Nomination & Report Request started (22 Jul 2019) by Jürg Luterbacher

RR by Lea Schneider (13 Aug 2019)

RR by Anonymous Referee #3 (13 Sep 2019)

Suggestions for revision or reasons for rejection

The manuscript has improved from its original form, but the authors have not fully addressed my concerns. The issue of spatial sampling biases has not been fully addressed, while several additional issues arise. I also list a collection of minor issues that must be addressed before the paper can be published. My recommendation is to send it back to the authors for another major revision based on my comments below.

The sampling bias has simply not been addressed. The bootstrap sampling that the authors perform does not evaluate the differences in the spatial sampling of the two proxy populations. It doesn't matter how many subsamples the authors use if they are drawing from two populations that are inherently biased toward two different geographic distributions. It would be much more useful to simply have a scatter plot of the proxy trends as a function of latitude. The data should be color-coded for the two populations (dendro and other), thus allowing the reader to see the range and density of latitudinal sampling and whether there is a trend dependence on latitude evidenced in the two populations. Such a plot is necessary and should be included among the results that the authors present. As I indicate below, the percent of records with significant trends is also not a very useful statistic when considering populations with sampling biases. Binning the trends in 5-degree latitude bands and then calculating the trend in the trend vs. latitude data for the two populations may be much more useful.

A new issue that emerges after reading the paper again is that the authors equate the presence of trends with low-frequency variability. These two things are not equivalent and the authors should be careful in their discussion. Spectral analyses should be included if the authors want to make the kind of definitely statements that they do about the expression of low-frequencies in the two populations. I also am surprised that the interpretation of trends is tied entirely to orbital forcing. The LIA is not mentioned at all, and will have a control over trends, particularly those computed over the last 800-1000 years. The expression of the LIA was heterogeneous and therefore adds to the importance of spatial sampling in the two populations considered. The LIA should be mentioned and the authors should explain why they feel comfortable interpreting trends exclusively through the lens of orbital forcing and not with a consideration of the LIA. Incidentally, given the prevailing focus on volcanism as a cause of the LIA and the associated hemispheric and latitudinal expressions of the many eruptions during the LIA, this is really something that should be given consideration in the manuscript.

A couple hold overs from the first review:

(1) Pg. 2, Ln 28: The list of multiproxy reconstructions does not include the 330 data assimilation work
(e.g. Hakim et al. (2016) and Steiger et al. (2018)) nor does it include the PAGES products from
2013 and 2018. This should be corrected.
(3) Corrected.

This was actually not corrected. These references are not included at the location in the manuscript discussed in this comment.

(1) Pg. 1, Ln. 13: It should be noted here that the 692 proxies are the temperature sensitive
records in the database (the full database is closer to 3000).
(2) No, the database includes exactly 692 records. We refer to PAGES 2k Consortium: A global
multiproxy database for temperature reconstructions of the Common Era, Nat. Sci. Dat., 4, 1-
320 33, 2017.
(3) No changes.

The authors are correct. Apologies for my oversight.

Pg. 2
Ln 33: defensible not defendable
Ln 41: from: not from;
Ln 43: no comma after tree/borehole (1)
Ln 54: A flat trend is the absence of a trend. Flat trend is nonsensical.

Pg. 3
Ln 65: Why would a globally averaged estimate of temperature be expected to have the same negative trend as a record from northern Scandinavia where the orbital forcing would be expected to be maximized. There is a stark contrast between two such records only in as much as there would indeed be a trend difference, but for entirely expected reasons. I don't think such a comparison and contrast should indicate anything significant about the fidelity of the underlying tree-ring records.

Pg. 5
Ln 35. How is the normalization of records with different resolutions handled? This makes it sound like records were first normalized and then averaged in 50-yr bins. But a normalized record with annual resolution will effectively be in different standardized units than a normalized record with decadal resolution. This is concerning given the nature of the comparison the authors are making, given that the dendro records will all have annual resolution, while the other records will have lower resolutions. The authors should confirm that they first binned and then normalized and not the other way around.

Ln 50: I do not understand the description here and it raises another concern with the analysis. I thought that all records of at least 800 years were selected and all trends were measured over the 800-yr period. But this sentence makes it sound like trend calculations were performed over variable time periods, i.e. some 800 years and others longer. If that is the case, the trends are hard to compare and may be influenced strongly by events prior to the 800-year window (especially when considering regression leverage) in a way that trends calculated only over 800 years are not. The authors need to clarify this and I would strongly argue for a common analysis window over which all of the trends should be calculated.

Ln. 55: (in-) significant cooling/cooling as a formulation is impossible to follow. This sentence makes no sense to me.

Pg. 7
Ln 01: It is these kinds of statistics that are hard to parse when there is uneven spatial sampling. Imagine the extreme case in which all the dendro records were in the tropics (admittedly impossible) and all the other records were from the Arctic. In such a case 100% of the high latitude records should have significant trends while 0% of the drendro records should have trends. Reporting such statistics would say nothing about the fidelity of trends in the records and have a clear physical explanation. While the divisions are not as strong here, I am simply unclear on how to interpret these percentages when the sampling biases exist as they do.

Pg. 9
Lns 67-69: I do not think this statement about internal variability is supported by the authors' results. Are they really suggesting that internal variability explains the differences in the records they have analyzed?

Hide

ED: Reconsider after major revisions (19 Sep 2019) by Jürg Luterbacher

AR by Lara Klippel on behalf of the Authors (29 Dec 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (12 Jan 2020) by Jürg Luterbacher

RR by Lea Schneider (16 Jan 2020)

RR by Anonymous Referee #3 (29 Jan 2020)

Suggestions for revision or reasons for rejection

This iteration of the manuscript is marginally improved over the previous version and I will not stand in the way of its publication any longer. There are some interesting analyses and various points worth including in the public literature. I nevertheless believe that the case the authors make is much weaker than some of their statements throughout and in some cases contradictory. The abstract, in particular, does not seem to have been updated based on the changes in the manuscript. It reads much more definitive in terms of the alleged absence of long-term trends in the tree-ring chronologies. For instance, the abstract states that during the 1-1800 CE interval "tree-ring proxies lack any evidence of a significant pre-industrial cooling" while the discussion states that "our results show that millennial-scale trends in NH proxy records are consistent between tree-ring, glacier ice and marine and lake sediment records when considering the period 1200-1800 CE to calculate the slope of pre-industrial trends but inconsistent between tree-rings and other proxies over the entire Common Era." I find the second statement much more qualified and not at all consistent with the definitive statement in the abstract that focuses only on the results from the 1-1800 interval.

It is also worth noting that the new Figure S6c makes clear what I have been saying all along, which is that it is hard to discern a latitudinal dependence in the pre-industrial cooling in either of the subset populations (if there is a dependance, it seems to be a decrease in the cooling trends with latitude!), a fact that is hard to resolve if orbital forcing were truly the explanation for the trend. Moreover, it is hard to see how one might conclude that significant differences exist between the tree-ring population and the other proxies. It is simply a fundamental figure that does not lend a lot of support to the authors' arguments. But Figure 5 also makes clear that not only does the 1200-1800 period produce similar trends across the dataset, even the 1-1800 assessment produces results in the very small population numbers that are quite similar, barring significance tests.

With all of the above said, the authors simply need to couch more of their conclusions about trends in the various proxies in caveats qualifying the uncertain nature of their results. Their work opens up some interesting questions about the representations of long-term trends in the various proxy data types, but I would not bet the proverbial farm on the robustness of the differences in the long-term trends that they have characterized. The manuscript should be revised with those qualifications in mind.

I also include a few small details in need of correction below:

Pg. 5, Ln 46-49: I cannot make sense of what binning approach the authors actually used here. Please be specific about what the PAGES2k method was (they say something about it in an earlier paragraph, but the details are vague). It is also not clear why they need to adhere to the PAGES2k method, given the concerns they raise in the previous sentence. One wonders why and the argument given is not a strong one.

Pg. 6, Ln 56: Milankovitch theory seems more appropriate than cycles here.

Pg. 12, Ln 63-64: It is rather rich sauce to first call in question the PAGES2k database and to have cautioned against the use of multiproxy reconstructions derived from that database and then to use the Neukom et al. (2012) paper, a study based entirely on PAGES2k multiproxy reconstructions, to further an argument here. Rich sauce indeed.

Pg. 13, Ln 77: detrending. not ,

Pg. 13, Ln 79: effect not affect

Hide

ED: Reconsider after major revisions (31 Jan 2020) by Jürg Luterbacher

AR by Lara Klippel on behalf of the Authors (17 Feb 2020) Author's response Manuscript

ED: Publish subject to technical corrections (11 Mar 2020) by Jürg Luterbacher

AR by Lara Klippel on behalf of the Authors (18 Mar 2020) Author's response Manuscript

Short summary

The PAGES2k multiproxy database offers a new and unique opportunity to study the lack of long-term cooling trends in tree-ring data, which can be expected in Northern Hemisphere summers, particularly in the high latitudes, due to orbitally driven changes in solar irradiance. Tests of different influencing factors reveal that preserving millennial-scale cooling trends related to orbital forcing is not feasible in most tree-ring datasets.