cc: Dith Stone <stoned@atm.ox.ac.uk>, Peter Stott <stott.peter@googlemail.com>, Toru Nozawa <nozawa@nies.go.jp>, Alexey Karpechko <A.Karpechko@uea.ac.uk>, Michael Wehner <MFWehner@lbl.gov>
date: Thu Aug 14 14:26:16 2008
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re: Fwd: Decision on Nature Geoscience manuscript NGS-2008-07-00710
to: Gabi Hegerl <gabi.hegerl@ed.ac.uk>, Nathan Gillett <n.gillett@uea.ac.uk>

    Nathan,
       I'd pursue this. If you can get it into Nature then all the better,
    although the resubmission to Nature would have to be well argued
    in the accompanying letter as well as in the text. The spatial
    issue is important in the Antarctic and the region was the
    one missing from AR4.

    Cheers
    Phil

   At 13:17 14/08/2008, Gabi Hegerl wrote:

     Hi Nathan, sounds like you got the foot quite well into the door if we can address
     reviewer
     1. I agree that the discounting of area results by lack of significance in individual
     stations is frustrating.
     Maybe though if one can give a bit more space to an explanation of why Antarctica does
     what it does
     would help (eg the SAM subtracted map)? Did the Monahan paper use the same kinds of
     models (with ozone forcing)?
     Gabi
     Nathan Gillett wrote:

     Hi all,
     We now have the reviews back on the polar temperature paper.
     Unfortunately it's rejected with a suggestion that we resubmit. One
     reviewer (reviewer 2) was very positive, and had few suggested
     changes. The other reviewer was unconvinced of the Antarctic analysis
     - his primary objection seemed to be that we shouldn't be able to
     detect anthro influence on Antarctic temperature if station
     temperature trends are not locally significant. However, he appeared
     not to consider that a large scale mean, or pattern of temperature
     trends may be significant even if individual station trends are not.
     Addressing these comments by calculating the significance of area mean
     temperature trends etc should be relatively straightforward - we've
     got to try to convince the non-specialist that the Antarctic trends
     are significant independently of the D&A analysis. I think it's worth
     revising and resubmitting to Nature Geoscience. Let me know what you
     think and suggestions for revising the paper.
     Cheers,
     Nathan
     ---------- Forwarded message ----------
     From:  <a.newton@nature.com>
     Date: 2008/8/13
     Subject: Decision on Nature Geoscience manuscript NGS-2008-07-00710
     To: n.gillett@uea.ac.uk
     13th Aug 2008
     *Please ensure you delete the link to your author homepage in this
     e-mail if you wish to forward it to your co-authors.
     Dear Dr Gillett
     Your manuscript entitled "Attribution of polar warming to human
     influence" has now been seen by 2 referees, whose comments are
     attached. Although they find your work of some potential interest,
     referee 1 has raised concerns which in our view are sufficiently
     important to preclude publication of the work in Nature Geoscience, at
     least in its present form.
     If, after future work, you can provide compelling evidence for the
     statistical significance of your reported Antarctic temperature trends
     as well as for your attribution of those trends to natural and
     anthropogenic forcing, we will be pleased to consider a revised
     manuscript (unless, of course, something similar has by then been
     accepted at Nature Geoscience or appeared elsewhere).
     I should stress, however, that we would be reluctant to trouble our
     referees again unless we thought their comments had been addressed in
     full, and we would understand if you preferred instead to submit your
     manuscript elsewhere. In the meantime we hope that you find our
     referees' comments helpful.
     Yours sincerely,
     Alicia Newton
     Associate Editor
     Nature Geoscience
     Nature Publishing Group
     The Macmillan Building
     4 Crinan Street
     London N1 9XW
     UK
     PS Please use the link below to submit a revised paper:
     [1]http://mts-ngs.nature.com/cgi-bin/main.plex?el=A1Q3CGj2A3FlJ1J7A93rUshYh3SdI0gPrnGWsf
     3wZ
     *This url links to your confidential homepage and associated
     information about manuscripts you may have submitted or be reviewing
     for us. If you wish to forward this e-mail to co-authors, please
     delete this link to your homepage first.
     +44 20 7833 4000
     Reviewers' comments:
     Reviewer #1 (Remarks to the Author):
     Summary:
     This paper attempts to formally attribute polar warming in both
     hemispheres to anthropogenic forcing. The approach of comparing GCM
     simulations forced by both natural (NAT) and natural + anthropogenic
     (ANT) has been used successfully in other attribution studies, but
     here the authors apply it to the polar regions where very little data
     is available. To isolate the difference between NAT and ANT, they
     employ an innovative detection and attribution technique.
     With regards to the Arctic, the model ANT trends appear to be
     reasonable compared to observations (Fig. 2). Due to the strong
     warming in the Arctic it would be hard to quarrel with the results for
     that region.
     With regards to the Antarctic, where there is less data and less
     warming than in the Arctic, the results are unconvincing. The authors
     fail to comment on a closely-related recent paper that has first-order
     consequences for their analysis (Monaghan et al. 2008b: 20th century
     Antarctic air temperature and snowfall simulations by IPCC climate
     models. Geophys. Res. Letts., 35, L07502, doi:10.1029/2007GL032630).
     That paper strongly suggests that the GCMs are too sensitive to
     anthropogenic forcing (more details below), and a key assumption of
     this study is that the GCMs are able to reasonably simulate
     anthropogenic influences on surface temperature. With this assumption
     in question, which is confounded by large uncertainty in the observed
     trends from the handful of available stations, and the fact that
     Antarctic warming is only likely to be statistically significant over
     a very small fraction of its surface area (<10%), the result that
     Antarctic warming is due to human influence is
     highly questionable. Detailed comments on the Antarctic analysis are
     given below.
     Without a convincing Antarctic analysis, I don't feel that this paper
     is suitable for publication in Nature Geoscience. Given the complexity
     of the topic, the authors might consider revising and submitting this
     important work to a high-profile journal that has room for a much more
     detailed analysis to be presented (Journal of Climate comes to mind).
     +++++++++++++++++++++++++++++++++++
     Detailed Comments on the Antarctic analysis:
     The authors seem to ignore the fact that there has been no
     statistically significant warming over 90% of Antarctica. For example,
     Figure 3 is misleading, since it does not show statistical
     significance. According to the statistics on Gareth Marshall's
     Antarctic temperature website ([2]http://www.antarctica.ac.uk/met/gjma/),
     and confirmed in Turner et al. (2005), the long-term annual positive
     temperature trend in the Ross Sea at Scott Base is statistically
     insignificant [1958-2007 trend = +0.0172 {plus minus} 0.0230], as is
     the long-term annual positive temperature trend at Casey near 110 E
     [1958-2007 trend =+0.0118 {plus minus} 0.0212]. Therefore, only over
     the Antarctic Peninsula (<5% of Antarctic surface area) and at a
     single station on continental Antarctica (Novolazarevskaya), has
     long-term statistically significant warming been recorded. The spatial
     influence of Novolazarevskaya appears to be very limited, as the
     stations on either side of it have statistically
     insignificant temperature trends near zero. So, the stations with
     statistically significant warming likely represent a very small area
     of Antarctica (<10%), and additionally they are sandwiched within a
     sector that only spans 80 degrees of latitude, from -68 W to 12 E. The
     authors try to rectify this localized warming by noting that "positive
     trends predominate" over most of Antarctica according to the surface
     temperature synthesis of Monaghan et al. (2008). However, Monaghan et
     al. (2008) noted that the positive and negative trends over Antarctica
     were overwhelmingly statistically insignificant apart from the
     Peninsula and a small region around Novalazarevskaya. Chapman and
     Walsh (2007) also performed a gridded Antarctic surface temperature
     reconstruction like Monaghan (over a longer period) and got similar
     results. The point is, how can the authors attribute Antarctic surface
     warming to anthropogenic forcing, when there is so little evidence for
     warming to begin with? Why
     didn't the authors use the more spatially-comprehensive data of
     Chapman and Walsh (2007) or Monaghan et al. (2008) for their analysis,
     even if just for comparison with their station-based results?
     Considering the distribution of the Antarctic stations and their
     comparatively short records with high interannual variability, perhaps
     the only place on the continent where one could argue for a robust
     anthropogenic surface warming signal is over the Peninsula. Marshall
     et al. (2006) made a convincing case that summer warming on the east
     side of the Peninsula is due to increased foehn wind events resulting
     from a stronger SAM. In turn, the link between the SAM and
     anthropogenic influences, especially from stratospheric ozone
     depletion, has been established in previous modeling studies, some of
     which the author cites here. Therefore, if one infers from the
     existing literature that the small region of Antarctica that has
     warmed statistically significantly during the past 1/2 century has
     been mainly influenced by the SAM, then the results shown in this
     paper for Antarctica (attributing surface warming to human influence)
     are not particularly groundbreaking.
     One key assumption of this study is that the AR4 models are able to
     accurately simulate the impact of anthropogenic forcing on Antarctic
     surface temperatures. However, in a very closely related study that
     was not cited in this analysis (Monaghan et al. 2008b), the authors
     found that 5 AR4 models, two that were included in this study, had
     annual surface temperature trends that were substantially larger than
     observed during the past ~1/2 century. The statistically insignificant
     observed 1960-99 trend from Monaghan et al. (2008b) was 0.06 +/- 2.03
     K, versus a highly significant GCM ensemble trend of 1.44 K +/- 0.34
     K; all 5 GCM members had statistically significant positive trends
     (p<0.05). The authors, who also compared their results to the 100+
     year Antarctic temperature record (1900-1999) of Schneider et al.
     (2006, GRL), found that the models results were much larger than
     observed over the past century as well. They examined why the GCM
     trends were so much more positive than
     observed and found (as the authors note in this paper) that the
     surface temperature sensitivity to the SAM is weaker than observed.
     More importantly, they concluded that in the GCMs, the influence of
     the SAM on surface temperatures appears to be overwhelmed by a
     spurious water vapor feedback. In turn, the water vapor feedback may
     be (wrongly) causing the much larger than observed GCM surface
     temperature trends over Antarctica. Their results indicate that the
     IPCC AR4 GCMs may not yet be able to fully simulate all of the impacts
     of anthropogenic forcing in Antarctica. If correct, their results
     signify that the key assumption of this study is not robust for
     Antarctica. Additionally, their study suggests that the Antarctic
     surface temperature datasets that are representative of surface
     temperature over the entire continent may yield a very different
     comparison with GCM results than is concluded from the comparison with
     the limited dataset used here.
     Reviewer #2 (Remarks to the Author):
     Overall this is an excellent manuscript and an important contribution
     to the detection and attribution debate.
     Detection and attribution studies require both models and
     observations, and this is often accomplished by comparing observations
     of actual changes to model-induced trends for models forced
     independently by natural, anthropogenic and combined forcings. This
     was first done, I think (Stott as an author will know for sure), by
     Stott, P., Tett, S., Jones, G., Allen, M., Mitchell, J. & Jenkins, G.
     (2000) Science
     290, 2133-2137.
     Stott, P. (2003) Geophys. Res. Lett. 30, 1728.
     using four ensembles of a single model, and very strong evidence for
     global-scale detection and attribution was offered and was a key
     element in IPCC TAR--the authors of this submission might make this
     history a bit more prominent in a minor revision.
     Another study with more limited data coverage arguing that some
     regional skill was still evident in the same set of model runs was
     offered using observations of spring phenology of plants and animals
     as a proxy for spring temperature, and again a clear detection and
     "joint attribution" to anthropogenic causation--though a smaller
     fraction of variance explained--was also found in:
     Root, Terry L., Dena MacMynowski, Michael D. Mastrandrea, and Stephen
     H. Schneider, 2005: "Human-modified temperatures induce species
     changes: Joint attribution, " Proceedings of the National Academy of
     Sciences, 102, 21, 7465-7469
     The latter used more sparse observational data and thus finding less
     variance explained than for global scale thermometer data in Stott et
     al papers was not surprising. BUt it did find skill at regional
     scales.
     In this submitted paper studying polar regions the authors aggregate
     four models, rather than one, and like earlier studies compare this
     for models driven by N and N&A forcings. Data in the polar regions is
     very sparse--more so than even the phenological ecological data
     sets--nevertheless the authors are admirably able to perform a
     heroic--and to me credible--effort to extract a signal of
     human-induced climate changes in this limited data set.
     My only suggestion to the authors is to consider framing their efforts
     in the context of earlier ones like mentioned above issues such as
     data coverage and show the evolution of D&A studies using N and N&A
     forced models and how all such studies at global to regional scales do
     agree that joint attribution is indeed a credible conclusion--and this
     latest study extends that to polar regions.
     In short, the authors should be congratulated on a fine addition to
     the literature.
     This email has been sent through the NPG Manuscript Tracking System
     NY-610A-NPG&MTS


     --
     Dr Gabriele Hegerl School of GeoSciences
     The University of Edinburgh
     Grant Institute, The King's Buildings
     West Mains Road
     EDINBURGH EH9 3JW Phone: +44 (0) 131 6519092, FAX: +44 (0) 131 668 3184
     Email: Gabi.Hegerl@ed.ac.uk
     The University of Edinburgh is a charitable body, registered in
     Scotland, with registration number SC005336.

   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
   ----------------------------------------------------------------------------

