Temperature variability in the Iberian Range since 1602 inferred from tree-ring records

Tejedor, Ernesto; Saz, Miguel Ángel; Cuadrat, José María; Esper, Jan; de Luis, Martín

doi:https://doi.org/10.5194/cp-13-93-2017

Articles | Volume 13, issue 2

https://doi.org/10.5194/cp-13-93-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/cp-13-93-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 13, issue 2

Research article

|

09 Feb 2017

Research article |

| 09 Feb 2017

Temperature variability in the Iberian Range since 1602 inferred from tree-ring records

Ernesto Tejedor, Miguel Ángel Saz, José María Cuadrat, Jan Esper, and Martín de Luis

Download

Final revised paper (published on 09 Feb 2017)
Preprint (discussion started on 02 Feb 2016)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Comments on cp-2016-9', Anonymous Referee #1, 12 Feb 2016
- AC1: 'Comments to referee1', Ernesto Tejedor, 25 Feb 2016
  - AC3: 'New version of the manuscript', Ernesto Tejedor, 25 Feb 2016
AC2: 'Comments for the referees', Ernesto Tejedor, 25 Feb 2016
- AC4: 'New version of the manuscript', Ernesto Tejedor, 25 Feb 2016
  - RC3: 'Response to authors', Anonymous Referee #1, 08 Mar 2016
    - AC6: 'Response to referee#1', Ernesto Tejedor, 28 Mar 2016
RC2: 'Review', Anonymous Referee #2, 25 Feb 2016
SC1: 'Comment on cp-2016-9', Stefan Klesse, 01 Mar 2016
- AC5: 'Reply to Stefan Klesse's comments', Ernesto Tejedor, 28 Mar 2016
SC2: 'Climate signal should be clarified', Chris Salt, 09 May 2016
- AC7: 'Response to Chris Salt', Ernesto Tejedor, 10 May 2016
RC4: 'General contents', Anonymous Referee #3, 06 Jul 2016
- AC8: 'Response to referee#3', Ernesto Tejedor, 12 Jul 2016
EC1: 'Editor comment', Volker Rath, 08 Aug 2016

Peer-review completion

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

ED: Reconsider after major revisions (08 Aug 2016) by Volker Rath

AR by Ernesto Tejedor on behalf of the Authors (29 Aug 2016) Manuscript

ED: Referee Nomination & Report Request started (22 Sep 2016) by Volker Rath

RR by Anonymous Referee #1 (17 Oct 2016)

Suggestions for revision or reasons for rejection

I am aware that a substantial amount of effort and time has gone into this manuscript and this makes harder to be so critical. However, and being honest, I still have serious doubts about the validity of the reconstruction presented in this paper. Trees are living organism and statistical analyses are supposed to be used to prove or reinforced an “a priori” hypothesized physiological relationships. The way the manuscript is presented and written gives the impression that the authors were looking for the highest correlation to develop a climate reconstructions and after finding it, they tried to physiologically justify the selection done. The result is a temperature reconstruction that seems statistically valid (with some issues as well, see comments below) but not reinforced by other analysis (i.e., SEA) and physiologically quite questionable, rising serious doubts about its trustfulness and validity.
I know that part of the problem is the over-interpretation and/or overestimation of the calibration statistics that has been done in some publications presenting tree-ring based climate reconstructions. However, the issue here is that if this work is published, the reconstruction will be “stored” in literature and data repositories as a temperature reconstruction and might be used in comparisons/ studies at regional scales. Therefore, it is compulsory to be completely sure that the climate signal encoded in the tree rings has not been misinterpreted and (unfortunately) I am not convinced about that:

1. I am familiar enough with tree growth in the Spanish territory to seriously doubt that a forest at mean of 1,758 masl in a continental sub Mediterranean climate and with more than 90 frost days per year grows better in cold years than in warm years. It may grow better during mild summers (i.e., cooler summer temperatures) but a cold year implies not only lower temperatures during summer, but also during winter and spring. At 1500 m of altitude, the length of the growing season is already constrained by low temperatures and a colder winter/spring implies a reduced growing period and therefore, a smaller tree ring. A delay in the onset of the cambium division will not be compensated by a favourable summer, because the largest part of the tree ring (early wood) is produced during spring, while during summer, cells are mainly in maturation phases. Adding the cumulative 2-years (or 21-cumulative monthly mean) component makes things even more senseless. This does not mean that a delayed climate signal in tree rings is not possible (e.g., Dorado Liñan et al 2015), but the cumulative climate signal that the authors describe is not physiological supported by the existing knowledge/bibliography and no robust explanation is given.

2. Regarding the issue of the high and low frequency agreement, it would have been clearer to show the correlations with climate of a high-pass and a low-pass filtered versions of the BasPois chronology, since this is the one used later on for the reconstruction. However, the correlation with ArstanRES gives an approximation and confirms that the correlation in the high frequency (r=-0.39) is lower than the one for the full series (ArstanSTD; r=-0.56) and much lower than the one for BasPois (-0.78). This makes me wonder: Is the standardization method used the best preserving low-frequency or was the best introducing and artificial trend matching the trend of the 21-cumulative monthly mean series? Figure 5 is quite enlightening on this regard: when low-frequency variations are removed (ArstanRES), it seems that no significant correlation with temperature are found (cannot find the r=-0.39 in the plot). Correlations are only significant when trends are present and this can be considered a sign of spurious correlation.
I have similar concerns to those expressed by S. Klesse regarding the biases that the estimation of the pith offset form the ITRDB data may introduce in the final chronology. In fact, that bias may have been determinant for the final result.

3. I think the authors made a mistake interpreting the DW statistic. A value of 1.45 p<0.001 (P9, line 10-11) means that there is a significant positive autocorrelation. On the other hand, the estimation of the significance of the residual’s trend is missing. The RE seems to support the calibration-verification done (RE is a verification statistic, it should only be applied in the verification period). However, the visual mismatch in the high-frequency, the lack of a coherent physiological explanation and the SEA analysis that shows significant impact 3 years after the volcanic eruption points to a misinterpretation of the climate signal. Maybe the tree rings encode a temperature signal but does not seem to be a 21-cumulative monthly mean temperature.

Hide

RR by Anonymous Referee #4 (16 Dec 2016)

Suggestions for revision or reasons for rejection

I was not a reviewer of the first version and I have tried to evaluate this revised version in view of the previous interactive discussions and based on the author's response to the reviewers.
My opinion is that the authors have largely responded to the comments in the first version, but not totally. I also think that they could have done a little bit more to present and compare the new standarization technique to RCS, so that that the reader could get a more comprehensive impression of differences and advantages of the new method.
Reviewer #1 of the initial submission also indicated that the physiological explanation of the correlation between the chronology and the mean temperature of the previous 21 months could be more detailed and more strongly backed on previous published works. The authors have expanded the manuscript a bit on this, and I would say that as a preliminary explanation it is acceptable, but this reviewer also has the impression that the authors did not really want to fully address this issue.
This being said, this version still contains a relatively long series of minor or particular issues/errors/ that need to be amended or clarified, as I explained in the list below.

More important comments:

1. One of the advantages of the RCS standarization method is that it iis usually presenetd as the method that best retains low frequency variability. This study presents another variant of RCS, based on an alignment of trees according to their size instead as according to their age. The obvious question would be how this variant behaves in the low-frequency domain compared to the RCS method. There is barely a word on this in the manuscript. As the bare minimum, both chronologies should be shown together in a figure, comparing the amplitude of low-frequency variations.
Related to this comment, it is unclear to me how the trees are really aligned in the BasPois chronology. The study mentions 'the 'square of the basal area' as the independent variable, but Figure 4 indicates cm2 es the units of the square of basal area. I think this is an error, and that more logically it is the basal area (not its square) the independent variable. This would match the units in Figure 4. Otherwise, I do not see the rationale of using the square of the area (it is probably the square of the diameter of the ring ?)

2. The physiological explanation for a response of the chronology to the mean of the monthly maximum temperature in the previous 21 months is basically a bit of hand waving. Maybe there is at this point no other explanation, but the authors could just candidly say it.

More particular points

3. The IPCC report
Explain the acronym IPCC. All acronyms should be spelled out, even if they are obvious.

4. a high-resolution temperature reconstruction

high-resolution in time, I guess. The sentence in this context is unclear as it could refer to a spatially resolved reconstruction over this region

5. Iberian Range
explain where the Iberian Range is located, at least broadly. The map shown in Figure 1 is of low-quality (at least in this pdf file) and should be improved for publication

6. 'RE is a measure of shared variance between actual and estimated series '

This explanation of RE is not totally correct/specific. ( the correlation is also a measure of shared variance). RE is better defined a measure of the typical size of the errors relative to the typical size of variations relative to the calibration mean.

7. To transfer the TRW chronology into a temperature reconstruction a linear regression model
was used

This is too unspecific. There are many variants of linear regression. I guess that the authors have set temperature as dependent variable and the chronology as independent variable, and have used Ordinary Least Squares assuming gaussian independent errors to estimate the regression coefficient. All this information is needed to completely specify the regression model

8.with a gradual decline of the growth until the cambial of 450. Cambial age from 500 to 550

until the cambial age of 450 years, I guess. Cambial age from 500 to 6550 years (?)

9. Calibration of the four differently detrended mean chronologies reveals a highly negative
correlation with maximum temperatures

with maximum temperatures or with monthly mean of daily maximum temperatures ? Taken the sentence in the manuscript literally, the authors mean the maximum temperature attained within the 21-month window. I do not think this is what they really mean.

10. correlations with previous-year September (r = -0.39), and the ArstanSTD chronology
correlates at r = -0.56 with September and October temperature of the previous year with a
cumulative monthly mean of 21 months.

Here and everywhere in the manuscript, this way of specifying the calender window with highest correlations is confusing. I agree that it is not straight forward to explain, but the characterization of September temperature is misleading. It is clearly not the September temperature, but the mean (maximum) temperature over a 21-month window. Thea authors should find a better way of defining this quantity. In this partucular sentence, I would suggest something like ' the 21-month mean temperature centered in September or October'. An efficient way would be to define it in the Methods section, attach an acronym to it (e.g. T_21_Sept or T_21-Oct or similar) and subsequently use the acronym through out the manuscript.

11. 'and the first principal component explains about 35% of the variance'

Have the chronologies been standarized to unit variance before the PCA ? otherwise the amount of explained variance is not informative, since it would depend on the indovidual variances of the chronologies

12. It is remarkable that the 12 years of the XXI century

In a scientific text, in English it is usual to refer to the 21st century, but this may be a mater of taste. Please, check.

13. ' The year-to-year temperature variability is ..'
The reconstructions refer to a 21-month calendar window, so it is confusing to refer to the year-to year variability. This is actually not clearly resolved in the chronology. Perhaps better refer to high-frequency (biennial) variability

14. The main driver of the large-scale character of the warm and cold episodes may be changes in
the solar activity

This point can be quite controversial and it is definitively contrary to present understanding, which states that the most important forcing for the midlatitude temperature variations is the volcanic forcing. It also seems speculative in this text since the authors have not conducted any proper attribution test to separate solar and volcanic forcing, which are known to be strongly correlated.

15. Overall, the correlation between the reconstruction and the solar activity is 0.34
(p < 0.0001), and increases to r = 0.49 after 11-year low pass filtering the series, thought the
degrees of freedom are substantially reduced due to the increase autocorrelation.

'though - > though
Another comment is that the number of degrees of freedom affects the statistical significance but not the magnitude of the correlation. A lower number of degrees of freedom does not per se on average artificially increase the correlation

16. The SEA (Fig.10) indicates some impact of volcanic eruptions on the short-term temperature
variability within the reconstruction. It shows significance (p < 0.05) decrease in September’s
temperature with a lag of three years.

The details of the SEA are obscure. The manuscript does not indicate which eruptions have been considered and how the significance has been established. This definively needs a longer explanation

17. developed a 410-year maximum September temperature reconstruction

developed a reconstruction of the monthly mean of daily maximum temperatures

18. signal to noise ratio, captures the regional climate signal accurately

a chronology never captures the climate signal accurately

19. In fact, climate variability is more size-dependent than age or
species (De Luis et al., 2009).

the impact of climate variability on trees may be more size-dependent, not the climate variability itself

20. Memory effects in TRW data have been also studied regarding the delayed response in TRW
(1∼5 years) to post volcanic eruptions associated with a decrease in current’s year
temperature (D’Arrigo et al., 2013, Esper et al., 2014)....... According to the SEA (Fig.9), the volcanic eruptions have a significance reduction (95% confidence) of September’s temperature (-1.98oC) with a three years lag.

This paragraph mixes two different effects, and it is not clear which one the authors are referring to. One effect is that the temperature response to eruptions is itself delayed, since volcanic aerosols need some time to spread globally in the stratosphere. The other effect is the physiologically delayed response of trees to sudden temperature drops ( or to reduction in sunlight caused by the eruption).

21. 2012 in agreement with the raise of temperatures observed for last decades

rise in temperatures

22. between the chronology and the climate parameter chosen never drops from -0.54

below -0.54

23. will trigger an incessant decrease in the tree-ring growth

would also cause a continuous decrease in tree-ring growth

24. Even though the CRU dataset extents the 1901-2013 period

The reader will wonder how is it possible that the CRU temperature records start in 1901 whereas the coverage of most meteorological stations starts in 1950. This would raise doubts on the quality of the pre-1950 temperature data

25. Even though the CRU dataset extents the 1901-2013 period, the general distribution of
meteorological observatories in Spain did not begin until the mid-twentieth century
(Gonzalez-Hidalgo et al. 2011)

spans the 1901-2013 period

26. However, based on a TRW
chronology, it is remarkable the high correlation coefficient for the full calibration period and
the CRU dataset (r = -0.78).

the high correlation coefficient is remarkable

27. However, previously to the Dalton minimum

prior to the Dalton Minimum

28. Overall, the correlation between the reconstruction and the solar activity is 0.34
(p < 0.0001), and increases to r = 0.49 after 11-year low pass filtering the series

which reconstruction of solar activity is being used here ?

29. Figure 4, caption

square of the basal area ? or basla area

30. Figure 5, Figure 7 caption

mean of daily maximum temperature

31 Figure 9
purple shading indicates the mean square error based on the calibration

I cannot see any purple shading in this pdf file

Solar forcing: which reconstruction of solar activity has been used here ?

32. Figure 10, how many eruptions and which eruptions have been used

Hide

ED: Publish subject to minor revisions (review by Editor) (16 Dec 2016) by Volker Rath

AR by Ernesto Tejedor on behalf of the Authors (26 Dec 2016) Author's response Manuscript

ED: Publish as is (18 Jan 2017) by Volker Rath

AR by Ernesto Tejedor on behalf of the Authors (19 Jan 2017)

Short summary

Through this study, and inferred from 316 series of tree-ring width, we developed a maximum temperature reconstruction that is significant for much of the Iberian Peninsula (IP). This reconstruction will not only help to understand the past climate of the IP but also serve to improve future climate change scenarios particularly affecting the Mediterranean area.