Comments on “Prospects for dendroanatomy in paleoclimatology – a case study on Picea engelmannii from the Canadian Rockies”

General comments:

This manuscript investigates the potential of dendroanatomical features of high-elevation Engelmann spruce from the Canadian Rockies as paleoclimate proxies. In doing so, the authors develop the longest dendroanatomical dataset for North America. I feel this study very interesting as it not only provides a comprehensive assessment of a relatively new tree-ring proxy type, dendroanatomy, but also attempts to compare dendroanatomical parameters with those obtained using the existing techniques - X-ray densitometry and blue intensity. I also appreciate that the authors test for the resolution related biases in BI and X-ray density and use two climate datasets to avoid biased results. This study finds that maximum radial cell wall thickness and anatomical MXD are the two most robust proxies of summer temperature and may be superior to MXBI and even X-ray MXD, where the latter has long been recognized as the most temperature-sensitive tree-ring proxy. These promising results will encourage dendroclimatologists to use anatomical parameters to better understand regional and large-scale climate history spanning centuries to millennia.

Overall, this study is valuable, although there are a couple of weak parts, that are outlined below.

Major comments:

My primary comments are about the long-term trend analysis in section 3.3. In dendroclimatology, it is widely accepted that many tree-ring parameters, even including dendroanatomical ones (see fig. S7 of Björklund et al., 2020), exhibit age effects which may mask true low-frequency climate variability embedded in tree-ring data. Thus, detrending is fundamental to remove these non-climatic, long-term trends, while it may also be challenged in specific cases. Indeed, it may be difficult to disentangle the long-term climate and age signals in even-aged samples as is the case of this study. Here, the authors try to interpret long-term trends of different parameters based on non-detrended series. Though “a robust picture of long-term trends needs RCS-type detrending” is acknowledged on Lines 578–579, I still wonder whether the analysis of the long-term trend based on the non-detrended data (mixed with both age-related and climatic trends) is appropriate. Moreover, the use of instrumental temperatures here to compare with non-detrended tree-ring data is confusing as non-detrended tree-ring data contain more information than climate.

While the authors’ analysis on the “high-pass” time series is good enough (not necessary to change), I would suggest using the detrended dataset for discussing the long-term trends (Lines 578–600), unless reasons and meanings of not using detrending could be sufficiently justified in the manuscript. The signal-free approach (Melvin and Briffa 2008) may be efficient to preserve more low-frequency climate signals. If the signal-free regional curve standardization does not lead to success due to low replication, signal-free approach plus age-dependent spline would likely be useful (see Wilson et al., 2019, Heeter et al., 2020, Wang et al., 2020). I also think even if these detrending methods may result in problematic long-term trends, it is still useful to present/discuss these chronologies rather than the non-detrended series. At least this will show that some tree-ring parameters are likely facing issues of retaining low-frequency signals, and further studies are thus needed. It might be useful to plot the age-related trend of each tree series for each parameter similar to (if necessary, replace) Figure 9, when discussing the success or failure of using these detrending methods. In addition, the reconstruction of Luckman and Wilson (2004) might be used as a reference to assess the long-term trends of detrended chronologies.

Additional comments:

Line 32: Long-term secular trends, please refer to the major comments.

Line 49: It is better to specify what are the “important climate periods”.

Lines 70–71: Perhaps also add Björklund et al., 2020 (see the recommended references in the end).

Line 80: should use “has become”...

Lines 97–102: Try to merge the two sentences. The second sentence is quite similar to the latter half of the first sentence.

Lines 104–105: Avoid saying that 15 trees are well-replicated. A well-replicated dataset may refer to a collection of hundreds of trees.

Line 109: X-ray radiography is not a new technique and Dendro2003 is a system designed about 20 years ago. There are now many advancing techniques providing higher-resolution density data, such as computed tomography (see Van den Bulcke et al., 2019). “the state-of-art” here should be removed.

Line 114: Be careful, en dash (–) should be used to represent the range in all cases, rather hyphen (-). Please check throughout the manuscript.

Line 115: What is the meaning of “pivotal locations” here? I think every location lacking a good proxy record can be pivotal for paleoclimatology. Consider removing “pivotal”.

Line 133: Try to move the arrow a little to indicate the correct location. A bit misleading here.

Line 136: Athabasca Glaciers is not indicated in Figure 1A). Please try to indicate it on the map, or remove the “Athabasca Glaciers” in the figure caption.

Line 145: It is better to use “immersed” instead of “washed”.

Line 158: The averaged cell wall thickness is not discussed in the manuscript. It is not necessary to keep it here. CWT should be moved to Line 157.

Lines 159–161: What is the tangential width of the measurement window? The 75^th percentile of what values? Are only the 75^th percentile values used or 0–75 percentile used? Please specify. In addition, by delimiting the 20 um bands some tracheids would be separated into different bands, I wonder how the CWT was obtained? It seems hard to measure “two radial and two tangential cell walls per tracheid cell” if tracheids are separated. Perhaps more details are needed so that the experiment could be repeated.

Line 179: Should indicate the version of CooRecorder.

Line 197: Figure S3 is not mentioned anywhere in the manuscript. Perhaps need to refer to Figure S3 here and re-order the supplementary figures.

Lines 204–205: The justification for not using RCS is not convincing enough. In my view, RCS works efficiently in cases where dead trees are also included, as it could avoid the “segment-length-curse”. If RCS does not work for the 15 living trees, the low sample replication and un-uniform growth patterns are more likely the reason. So, consider clarifying here that low replication hampers the use of RCS. I also suggest giving some supplementary graphics of RCS chronologies to explain why RCS is not suitable. In line with my major comments, it is worth trying signal-free RCS and signal-free age-dependent spline smoothing as well.

Lines 213–215: Is there a particular reason why the robust mean is not used here? Specify if possible. However, it is not a big problem. It is better to use “detrended data” (or “Spline smoothed data”, if a second detrending method is used; see major comments) in the manuscript because some medium-frequency signals may still be retained by using the 35-yr spline smoothing. “High-pass filtered data” sounds like all low frequencies are filtered out.

Lines 220–221: Please avoid writing “sufficient sample depth” here. The minimum sample depth is only 9, which even doesn’t lead to an EPS>0.85 for many parameters according to Table 1.

Line 229: Caution, the term “cross-correlation” is wrongly used in this manuscript. Generally, in statistics, the term “cross-correlation” represents the lagged correlations between time series, rather than correlations across time series. “Pairwise correlation” should be used in the entire manuscript.

Line 241: The correct citation should be “St. George and Luckman (2001)”. Change it also in the reference list.

Line 245: Citation or URL of Meteorological Service of Canada should be added. In addition, I couldn’t find the gridded data spanning 1895–present from the Meteorological Service of Canada. Where the data could be accessed?

Line 264: What is the time period used to calculate the r bar and EPS? The period 1585–2014 here is not consistent with Line 219 which describes 1700-1994 is used for the subsequent analysis. Please be consistent.

Line 266: If there is only one study cited here, the sentence should be: “with a previous study...”

Line 272 and Table 1: How the n for EPS = 0.85 (the last column of Table 1) is estimated? Some n is even greater than the actual number of samples. I guess “n for EPS = 0.85” is calculated based on the rbar and the equation of EPS. Please clarify this somewhere perhaps with an equation, e.g., at the bottom of Table 1.

Lines 301 and 414: See the comment for Line 229.

Lines 336–337: Should use “The first two components together represent 68.1% of the total variation”. Should also clarify what correlation is used, Pearson’s r?

Lines 339, 390, and 418: see the comment for Lines 213–215.

Lines 364–366: Perhaps also useful to compare the seasonal responsive window with Picea species in the North American continent. For example, black spruce in the eastern North American boreal forest, in similar latitudes.

Line 391: “Monthly or seasonally averaged temperature”.

Lines 411–413. Besides the low measurement resolution, the color sensitivity of the BI method may also affect the signal strength of MXBI to some extent. It is hard to ensure that all the measured wood cores are completely free of color biases. It is thus highly appreciated to use a few sentences somewhere to illustrate that weaker signal strength may also result from the color sensitivity of MXBI even for unstained living trees, see Wang et al., 2020. This is a fair discussion about the BI method.

Line 430: consider referring to Figure S3 after “in the widest rings”.

Line 439: r² should be used instead of r2. Sometimes R² is also used in the manuscript. Please keep the expression consistent.

Lines 485–487: Consider simplifying the sentence: “were further assessed by a split-period calibration procedure (1901–1948 and 1949–1994)”.

Lines 488–489: Refer to the major comments.

Lines 513: Since the r² distribution consists of 1000 values, it is better and more logical to perform some statistical test (e.g., two-sided student’s t) to show whether correlations of different parameters are similar or not.

Line 516: Figures 7–9 were mentioned only once each in the main manuscript. Maybe should consider moving them to the supplementary material. In addition, “correlated width” should be changed to “correlated with”.

Line 530, figure 6: the top-left annotations appear not complete. Only Max. radial CWT in A) and only aMXD in B)? Since aMXD and X-ray MXD are more directly comparable, why they are plotted in two separate plots? Perhaps show chronologies using either signal-free RCS or signal-free age-dependent spline here, after the major comments are taken into account.

Lines 550–611: please refer to the major comments.

Line 560: Should use “lower frequencies”.

Line 625, Figure 9: Why JJA temperature is used here? Ring width contains JJ signal and density parameters contain JA signal.

Line 655: Please remove “the state-of-art”.

The reference list: check the format on Lines 729–731; Line 745: “IAWA J”; Line 755: “Science”; Line 795: remove “Verona”; check Lines 804–806; Are Lines 829, 832, and 870 necessary? Line 864: volume and page numbers should be added.

References cited in this review:

Björklund et al., 2020. Dendroclimatic potential of dendroanatomy in temperature-sensitive Pinus sylvestris.

Heeter et al., 2020. Late summer temperature variability for the Southern Rocky Mountains (USA) since 1735 CE: applying blue light intensity to low-latitude Picea engelmannii Parry ex Engelm

Luckman and Wilson, 2004. Summer temperatures in the Canadian Rockies during the last millennium: a revised record.

Melvin and Briffa, 2008. A “signal-free” approach to dendroclimatic standardization.

Van den Bulcke et al., 2019. Advanced X-ray CT scanning can boost tree ring research for earth system sciences.

Wang et al., 2020. Temperature sensitivity of blue intensity, maximum latewood density, and ring width data of living black spruce trees in the eastern Canadian taiga.

Wilson et al., 2019. Improved dendroclimatic calibration using blue intensity in the southern Yukon.

The paper „Prospects for dendroanatomy in paleoclimatology – a case study on Picea engelmannii from the Canadian Rockies“ by K. Seftigen et al., explores the paleoclimatic potential of a broad set of dendro anatomical proxies, in particular focusing on the relationships between “new” proxies (mostly related to wood anatomy) and proxies that are relatively more known, such as X-ray maximum latewood density (MXD) and blue intensity (MXBI), and already used in the boreal environments of North America. My general impression about this manuscript is very positive, the paper deals with a timely issue, and such an assessment of the strengths and weaknesses of dendro anatomical proxies is a much-needed help for anyone working with “alternative” tree-ring paleoclimatic proxies. It is also notable that long chronologies of wood anatomical parameters are finally emerging even for the North American continent. For this reason, I can easily foresee that this work has the potential to become a highly cited reference in the field.

That said, I think there are still a few points that require attention.

Major issues

The first major issue is related to the chronology development of wood anatomical parameters: in lines 159-165 it is specified that wood anatomical parameters were calculated as the 75^th percentile within 20 µm wide bands parallel to the ring boundary. Nothing to say against this tree-ring partitioning approach, but it is not clear how the data obtained for each band were treated: usually, when the tree ring is partitioned, then multiple chronologies are developed. Since I do not see this in the manuscript, I assume that these data were somehow averaged. If so, it should be specified. In any case, the procedure of chronology development should be clearer, avoiding pointing to third papers.

Secondly, I have major concerns related to the procedure for testing temporal stability of dendroclimatic relationships in dendro anatomical proxies (Ln250 - Ln259 & Ln452 - Ln 473):

1) if the goal is to test the temporal stability of dendroclimatic relationships obtained for the monthly data, why shift from using monthly data to daily data, using a different climatic dataset (the Berkeley Earth dataset) that is, in addition, experimental and that You seem not to trust completely (see Ln. 461-464)? I think this is simply wrong and cannot be accepted unless a thorough comparison between the two datasets (CRU TS v4.03 and Berkeley Earth dataset) is performed to check if there exist substantial differences between the two.

2) if daily data are going to be used (and I am a huge supporter of daily data, whenever available, especially with wood anatomical data) why use them in 30-days fixed windows (so basically reconducting to a monthly analysis)? This is not the approach proposed by Jevsenak and Levanic (2018), which I would actually recommend following, testing moving correlations over the same 30-year period but using temporal windows starting from i.e. 14 days up to whatever length You'd like to test (30, 60, 180 days). That would help better target the temporal windows of climatic sensitivity described in Figure 3 and Figure 5.

A third major concern is related to the overall presentation of the paper. Despite being well written and accompanied with high-quality pictures, the manuscript is, in my personal opinion, exceptionally long, and at some point, allow me to use this term, becomes exhausting to read. I refer, for example, to the analysis related to the biases in the MXBI technique that are introduced in lines 190-199 and then discussed in lines 411-450. Honestly, I do not see the point of this long analysis and discussion. I am not criticizing its soundness nor questioning its interest. But to me, it looks like material for a stand-along paper, a deep technicality related to the blue intensity with weak (or at least not immediate) implications for dendro anatomy within the context of this manuscript. I believe that the strength of this work is that it is exploring important technical details related to those dendro-anatomical proxies that are less known in this region. It should not focus on going into deep technical details related to proxies or techniques that are already relatively more known and applied. Another example is in lines 368-385. This long paragraph discussing the climatic response of EW features could be easily avoided, since it is evident throughout the entire manuscript, and according to Table 1, that EW signal is not predominant at this location. Since the paper is already very long, this long (and sometimes repetitive) discussion on EW density and lumen area could be avoided. The last example could be section 3.1, where several times concepts discussed in the following sections are already mentioned. In general, I suggest careful evaluation throughout the manuscript if certain paragraphs and/or concepts could be removed or rephrased to improve the readability of this paper.

Minor issues:

The paper introduces the longest dataset of dendro-anatomical parameters for North America but there is not a figure showing such 1585-2014 chronology in the paper for any parameter. I suggest adding it, even if the analysis is focused on the 1700-1994 period. Moreover: why is the chronology truncated in 1994? And the data in Table 1 refers to the 1585-2014 or 1700-1994 timeframe? Please clarify.

Lines 320-331 Could be integrated into the next section.

Lines 360-361 “Even though the parameters describe two temporally distinct temperature signals, both are encapsulated within the short June-July-August period”. What are the two temporally distinct temperature signals? Please clarify.

Lines 363-364 “…average monthly temperatures rise above 0 °C only in four months of the year”. In Fig 1c it is actually five months.

Lines 364-366 “This window is substantially shorter than the single but wide target season observed in the latewood anatomical traits of P. sylvestris growing in temperature-limited environments in northern Scandinavia”. I don’t understand this sentence (i.e. what is the “target season observed in latewood anatomical traits”?), please be more clear.

Lines 368-385 Regardless of my previous comment (see above), I would like to add a few words about this. The reversing relationship observed between June-July temperature and EW density and lumen area might be due to an effect of precipitation, or in general of moisture availability, at that time. The initial stages of wood formation (hence, the anatomical features of the cells formed firstly along the ring) are highly dependent on water availability, which determines cell turgor in the enlargement phase. This is generally reflected in the EW lumen features, such as lumen area or lumen radial diameter. As a general remark, I would not say that EW features (i.e. lumen area) are not suitable climate proxies (or at least I would properly contextualize it): this might (is) true for this location and for temperature limited environments, but in arid environments earlywood (not latewood) features (in particular lumen size) are a crucial climatic proxy (not cell wall thickness).

Figure 4 The r coefficients listed on the right side, how are the aggregate months' correlation computed? Is it an average of the two r? Why some values are listed for both single and aggregated months, and in other cases not, even though the correlations seem equally strong?  

Lines 458-459 Please clarify the meaning of "peripheral ends" and of "elusive".

Lines 484-489 This is a key methodological aspect that lead to one of the main findings of this paper, hence I was surprised to find it here and not in the Methods section. Please move it to the Methods section.