date: Wed Jul  9 10:07:11 2008
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: [Fwd: Your Submission]
to: "Bo Vinther" <bo@gfy.ku.dk>, k.briffa@uea.ac.uk, ddj@gfy.ku.dk, hbc@gfy.ku.dk, kka@gfy.ku.dk, sigfus@gfy.ku.dk, bo@gfy.ku.dk

    Bo et al,
       I went onto the QSJ web site and found the comments from Rev 1, so
    am attaching them.
       Rev 1 says a lot, but all is positive and shouldn't take too long to include.
    Some of the necessary responses will make useful additions to the paper.
    A few may even have helped Rev 2. As for Rev 2, they also appear to want
    much more detail and much more justification for what has been done. They
    also seem to contradict themselves a couple of times. I'm much happier with
    fishing than relying on theory, so give me correlations any day!
       What would likely placate them the most would be plots with correlations between
    SW Greenland, Stykkisholmur and the long Angmassalik record on E Greenland.
    The length of the paper will likely increase, so it would be worth considering
    putting one or two parts into Appendices if this is possible.
    Cheers
    Phil
   At 02:24 09/07/2008, Bo Vinther wrote:

     Dear Phil, Keith, Dorthe, Katrine, Sigfus and Henrik
     I have received the reviews on our seasonal O18 manuscript for QSR. As far
     as I can see Reviewer 1 is very positive and has some suggestions for
     improvements while Reviewer 2 finds the paper too specialized and is
     unhappy with the statistics and in general sceptic of the results we
     present....
     The editor wants us to revise the paper - which I intend to do - but
     probably not before late August when I return from the field. Luckily that
     does not seem to be a problem.
     I will mail you all a draft revised version of the paper as well as a
     point by point response to the reviewers comments as soon as I have this
     ready...
     All the best
     Bo
     ---------------------------- Original Message ----------------------------
     Subject: Your Submission
     From:    "Quaternary Science Reviews" <JQSR@elsevier.com>
     Date:    Tue, July 1, 2008 10:53 am
     To:      bo@gfy.ku.dk
     Cc:      a.j.long@durham.ac.uk
     --------------------------------------------------------------------------
     Ms. Ref. No.:  JQSR-D-08-00106
     Title: Climatic signals in multiple highly resolved stable isotope records
     from Greenland
     Quaternary Science Reviews
     Dear Dr. Bo M Vinther,
     Reviewers have now commented on your paper. You will see that they are
     advising that you revise your manuscript. Both referees see important and
     novel work in your paper but raise some significant issues regarding the
     processing,interpretation and presentation of the data.  Although referee
     1 suggests that the paper is too specialised for QSR, I disagree and think
     that a suitably revised paper has the potential to make an important
     contribution to the journal.
     I ask that if you wish to revise your paper, you submit a detailed
     response to the specific comments raised by each referee.  It is likely,
     given the nature of the comments raised, that I will send the revised
     manuscript out for re-review.
     For your guidance, reviewers' comments are appended below.
     To submit a revision, please go to [1]http://ees.elsevier.com/jqsr/ and login
     as an Author.
     Your username is: bmv
     Your password is: vinther6573
     On your Main Menu page is a folder entitled "Submissions Needing
     Revision". You will find your submission record there.
     Yours sincerely,
     Antony Long
     Editor
     Quaternary Science Reviews
     Reviewers' comments:
     Reviewer #1: see attached pdf file
     Reviewer #2: The authors separate high-resolution isotope records from 20
     ice cores in Greenland (14 after stacking) into winter and summer
     timeseries, on the basis of correlations with Greenland and Iceland
     instrumental temperature records. They argue that the winter data are more
     spatially coherent than summer data, and have a stronger climate signal
     (despite more noise in winter). They then use 3 long records to extend the
     analysis back to ~600 AD. The most interesting result is the better
     correlation of winter isotope timeseries with borehole temperature
     inversions than summer and annual timeseries, but the differences are seen
     most readily before ~1300 AD. Curiously, the winter isotope data exhibit a
     millennial-scale cooling trend that is similar to the well-known NH
     temperature reconstructions, which are primarily derived from summer
     sensitive tree-rings. The authors do not discuss this last point. The
     writing is, for the most part, clear and well-organized.
     It is an interesting analysis, and I am impressed that the seasonal
     splitting can be done, yielding some promising results, especially the
     borehole-isotope comparison. Unfortunately, the analysis raises more
     questions than it answers, and has insufficient detail and explanation of
     methodology. Furthermore, the referencing and discussions are quite
     limited, and as such the manuscript does not seem appropriate for a review
     journal like QSR. I think the paper requires major revision before being
     re-submitted, perhaps to a more specialized journal. The following major
     issues should be addressed, followed by a number a minor issues.
     MAJOR ISSUES
     One) The limitations and assumptions of diffusion modeling. What
     parameters drive the diffusion model? How were these estimated? What is
     the sensitivity of the results to different choices of parameters?
     Two) On what physical basis is the 50/50 split between winter and summer
     isotopes made?
     What about the seasonality of precipitation? It is different in SW
     Greenland vs central and E Greenland? See Bromwich et al (1998, JGR 103,
     D20, 26007-26024).
     Three) Misleading interpretation of Figures 2 and 3.
     The authors have optimized the correlations of isotopes with a small set
     of temperature records. I think it is always dangerous to explain a
     variable like del 18-O in terms of a two-parameter model (SW Greenland and
     Stykkisholmur), without any regard to the physical processes driving the
     isotope signal. The model is bound to be over-fit (if only conceptually)
     by forcing the data into this simple 2-parameter framework; the analysis
     borders on "fishing" for correlation "bites."
     In figure 2, there are little differences in correlations for isotope
     fractions from 50-100%, whereas there are comparatively large differences
     in the choice of instrumental data, especially in panel b. In panel 3a,
     there is both a dependency on isotope fraction (improves as more data are
     included) and a dependency on how many months included in the temperature
     record (improves systematically from 2 to 12 months selected, especially
     in panel a). This improvement is not logical at face value, as the
     "summer" isotopes are poorly correlated with summer temperature (July-Aug
     and Jun-Sep curves are at the bottom in 3a!!), and better correlated as
     more winter data are included from the instrumental record. Furthermore,
     winter centering works better for SW Greenland, whereas summer centering
     works better for Stykkisholmur temperatures. Does this have something to
     do with the seasonality of precipitation?
     I would conclude first, that in most cases, including more than 50%
     isotope data leads to better correlations with temperature, and it doesn't
     hurt that much in the other cases (in panel 2a for example, where the max
     correlation only drops from ~0.68 to ~0.65 from 50% to 100%). In this
     regard, it is apparently the exclusion of winter data that hurt the
     correlations more than the inclusion of summer data. Second, it seems to
     me that the correlations have a stronger dependency on the selection of
     instrumental data than on the selection of isotope fraction. This is the
     opposite conclusion of the authors, who argue that the splitting of the
     isotope records fundamentally alters the interpretation of the records. I
     don't think our interpretations are altered that much, as we have known
     about NAO signals in Greenland isotopes for a long time (going back at
     least to White et al, 1997, JGR, 102 C12, 26425-26439), before any
     seasonal splitting was done. Furthermore, we cannot reach a
     conclusion about the merit of winter-centering vs summer-centering the
     year, as it depends on which instrumental record is used.
     Four) It is not explained how field correlations are derived from the 14
     point observations in Figures 7,8,9,10,13,14,15, and 16. How are these
     maps derived from 14 sites? There appears to be at most, 3 levels of
     correlation in each map. Why not show the correlation and regression
     coefficients for each ice core record?
     Fifth) The statistical significance of correlations is not assessed, and
     the authors use inconsistent criteria for discussing significance.
     Six) Are the winter and summer data really independent? Why do the PC
     patterns in Figures 10 and 16 look so similar? What is the correlation
     between PC1-summer and PC1-winter? It should be near zero, but it doesn't
     look like that is the case.
     Seven) The coherency among records is only partly assessed. It would be
     helpful to show the correlations among the 14 records. The differences in
     level of variance explained in summer vs winter PCs, and the correlations
     in tables 3 and 4, aren't different enough to convince me that summer
     records are less coherent than winter records. If summer is less noisy,
     shouldn't summer be more coherent?
     Eight) Incomplete referencing and discussion, especially since this is
     supposed to be a review article. Vinther and colleagues are not the first
     to evaluate the NAO and temperature signals in Greenland ice cores. I am
     surprised at the omission of White et al 1997, the more recent work of
     Schneider and Noone (2007; JGR, 112, D18105, doi:10.1029/2007JD008652),
     and others. There is no discussion of any physics, or of any modeling
     studies (e.g. Werner and Heimann, 2002, JGR 107 D1, 10.1029/2001JD00253)
     that might support (or not) the results.
     MINOR ISSUES
     Introduction, pg2: It's claimed that the NAO is the dominant mode of the
     NH. But the North Atlantic is not the entire Northern Hemisphere! An
     objective analysis of the entire NH SLP field reveals the
     hemispheric-scale Northern Annular Mode/Arctic Oscillation to be the
     dominant mode (e.g Quadrelli and Wallace, 2004, J Clim 17, 3728). The NAO
     can be viewed as part of this larger mode, but is not by itself the
     dominant mode.
     Data section 2.2: It is not explained how representative the Stykkisholmur
     record is of E Greenland temps.
     Diffusion section 2.1: It is not explained which parameters were inputs
     into the diffusion model, and how these were chosen.
     Seasons, section 3.3: The seasonality of precip is not addressed. You
     can't assume that del 18 O max/mins universally equal temp max/mins across
     the ice sheet, and that the max/mins occur at the same time of year
     everywhere.
     Pages 5-6: See my major comments above. The 50/50 split seems too
     arbitrary, and the claim of independence of summer and winter is not
     demonstrated. Are PCs1 of summer and winter independent? What is the
     physical basis of these correlations?
     In this and all following sections, statistical significance has been
     neglected.
     Section 4, pg 6: How do you define SNR?
     Pg 7, top paragraph: If lower variability is forced on the S Greenland
     isotope timeseries, why not do the same to the instrumental data to make a
     more straightforward comparison?
     Pg 8: If PCs 2 and 3 are statistical noise, how can you use them to reach
     a conclusion about the influence of the Central Greenland ice divide on
     air masses? This assertion is also not supported by the regressions on Fig
     12.
     Pg 9, section 6.1: I suspect that all of the correlations below 0.25 or so
     are not significant, so there is not really a clear SW-NE reduction in
     significance, just noise. The significance level also depends on the
     auto-correlation of the timeseries, which is not shown. So a lower
     correlation in the NE sector could be more significant than a higher
     correlation in the SW sector.
     Pg 10, section 6.4 and Table 4: It is not demonstrated that 34.9 % for
     summer and 39.9% for winter are statistically different, if you were to
     say, look at the error bars on the eigenvalues. If PC2 and PC3 are more
     meaningful in summer than winter, why do summer and winter loading
     patterns look virtually identical?
     Pg 11-12, section 7.1: Neglect to mention that the effective resolution of
     the borehole inversions decreases back in time, which may affect the
     correlations, and the ability of the borehole record to resolve the
     high-frequencies seen in the isotopes. Also, DYE3 borehole has larger
     amplitude variation than GRIP, which is not explained. What are the
     correlations among the timeseries in Figs 19 and 20, so that we can see
     that winter is clearly better? For DYE3, why do the annual and summer fits
     look pretty good for 1300-1970, but not prior to 1300?
     Pg 12: It might not be helpful to include summer isotope data to compare
     with winter and annual temps, but the exclusion of winter data to compare
     with annual temps is probably the bigger issue.
     Section 7.2: The discussion of Table 4 is difficult to follow. It appears
     to me the C/G correlations are better for summer and annual than for
     winter, the D/C correlation is similar across seasons, and the D/G
     correlation only works for winter. Why not use a C/G stack or a D/C stack
     that is good for all seasons?
     How do these stacks compare with PC1, and are they better/worse correlated
     with instrumental temperatures?
     How does a D/G isotope stack compare with stacked D/G borehole records?
     Section 8, pg 13: ".the summer data are much less influenced by noise and
     significant atmospheric pressure patterns can be identified, even down to
     the third summer del 18 O PC."  At what point in the manuscript are
     "significant atmospheric pressure patterns" shown for summer? In section
     6.5, if one cannot interpret PCs1-2, it is not valid to interpret PC3.
     TABLE1: It would be helpful to list the time span covered by the ice core
     records, and an indication of which records were stacked together. This
     would be more useful than the drill year & core length.

   Prof. Phil Jones
   Climatic Research Unit        Telephone +44 (0) 1603 592090
   School of Environmental Sciences    Fax +44 (0) 1603 507784
   University of East Anglia
   Norwich                          Email    p.jones@uea.ac.uk
   NR4 7TJ
   UK
   ----------------------------------------------------------------------------

