Friday, June 15, 2012


cc: Brazdil Rudolf <>, Reinhard Boehm <>, "'Petr Dobrovolny'" <>,, Juerg Luterbacher <>, Jucundus Jacobeit <>,
date: Tue, 17 Jun 2008 17:29:23 +0200
from: Petr Dobrovolny <>
subject: Re: AW: AW: Central European temperature reconstruction
to: Anders Moberg <>

Dear colleagues,

1) we made some cross-checking of GER, CH and CZ document. indices on
seasonal level only - I agree with Rudolf to stay on seasonal level only.

2) Average instr. CE series - my experience is - the more stations the
better. Of course it is necessary to do some mutual comparison. I can do
it next days with the set of stations suggested by Anders. So - I vote
for including Austrian stations also.

Best regards


Anders Moberg napsal(a):
> Dear Rudolf,
> cc. all
> Some quick responses:
> 1. Yes, the monthly series are on the discrete scale from -3 to +3 for
> each country. But - as we are averaging CH+DE+CZ, we actually have a
> scale from -9 to +9 even for the monthly series (in years when all
> countries have data, which is not always the case - but mostly we have
> data from at least two countries). I think this speaks for producing
> the monthly resolution. The paper can focus on seasonal resolution,
> but we could anyway produce monthly series and add as 'additional
> supplementary data'.
> 2-3. Yes, as we will use documentary data only for CH+DE+CZ, it is
> natural to use instrumental data also for these three countries.
> However, Austria nicely 'fills in' a geographical 'gap' between CH and
> CZ. In addition all the EI-corrections are based on Austrian
> measurements. This speaks for the use of CH+DE+CZ + AU for the
> instrumental series.
> 4. Yes, it is very good if Petr can construct homogenised Prague data
> separately for JFMA
> cheers,
> Anders
> Brazdil Rudolf skrev:
>> Dear Anders, dear colleagues,
>> it is really great that also other colleagues are now involved in our
>> discussions related to preparation of CE article for SI CC. There is
>> no doubt that possibility to work with homogenous data significantly
>> improves our possibilities and results. We spent with Peter a lot of
>> time trying to homogenise Prague and Rapp series and it was done on
>> seasonal and annual levels. Of course, in the situation we have also
>> some German data available it is better to calculate any new German
>> series and we can try to look once again on Prague data. But we
>> should not forget the original aim of our paper fo SI CC - i.e.
>> reconstruction of CE temperatures based on documentary and
>> instrumental data since 1500. In this connection we need instrumental
>> data for calibration/verification of documentary data and factual
>> extension of instrumental data to AD 1500. In connection with this I
>> propose following:
>> 1) for supposed article in SI CC present results only on seasonal and
>> annual level - working with documentary data on monthly level is
>> biased by higher uncertainties than in case of seasons (e.g.
>> resolution - for monthly indices you have only scale -3 to +3, but
>> for seasons -9 to +9, we did not cross-checked interpretations of
>> indices on the monthly level etc.)
>> 2) if we have documentary data only for D, CH and CZ, then
>> corresponding "CE" series should be really calculated only from
>> series of these 3 regions
>> 3) in case we would like to have real CE temperature series, then we
>> should include not only Austrian data, but also Polish, Slovakian a
>> Hungarian, how is usually Central Europe geographically taken
>> 4) as mentioned by Peter, we are able to homogenise monthly Prague
>> temperatures for JFMA for Juerg.
>> With best regards,
>> Sincerely
>> Rudolf
>> Anders Moberg napsal(a):
>>> Dear Reinhard, Rudolf, Petr, Phil, Juerg, Jucundus, Gerard
>>> First of all; thanks, Reinhard, for this additional info. You
>>> certainly don't need to convince me - I am already convinced that
>>> your EI-corrected versions are better than the non-corrected
>>> versions. This is certainly a step in the right direction.
>>> And good to have Phil onboard this little email conversation! It was
>>> quite a time ago since we did some real work together. Phil; to make
>>> a long story short - all you need to know at the moment, is that we
>>> (Petr, Rudolf and others) aim at developing temperature series (at
>>> least seasonal, but monthly are possible) for Central Europe back to
>>> 1500 based on documentary evidence from Switzerland (Pfister),
>>> Germany (Glaser) and Czech Republic (Petr, Rudolf), calibrated by
>>> means of linear regression against long instrumental records over a
>>> period overlap, and then using standard verification statistics plus
>>> some additional work to modify the error bars to take into account
>>> changes in expressed population signal over time. The calibrated
>>> documentary series will then be used to extend the instrumental
>>> record back in time. This is where Reinhard's new EI-corrected data
>>> come in!
>>> I also put Juerg, Jucundus and Gerard on this email, as a
>>> continuation of the discussions we had yesterday here in Ascona.
>>> Since yesterday, I had a look in the Rapp report from 2000. I just
>>> realized (and Petr just 'discovered' the same thing independently,
>>> as he just wrote that in an email to me) that Rapp's "German" series
>>> for the period 1761-1890 actually is an average of De Bilt, Berlin,
>>> Wien, Basel!! Obviously, we cannot and should not use the Rapp
>>> series for our current work. I suggest instead the following
>>> appropach, which is slightly different compared to what Petr, Rudolf
>>> and I have been talking about recently:
>>> We construct one temperature series per country (Switzerland,
>>> Germany, Czech, Austria), and then simply average the four country
>>> series together to form a Central European temperature series. Voil�!
>>> The CH component could be an average of Basel, Bern and Geneva.
>>> The DE component could be an average of Hohenpeissenberg, Karlsruhe,
>>> Munchen, Regensburg and Stuttgart.
>>> The AU component could be an average of Innsbruck, Kremsmuenster and
>>> Wien.
>>> The CZ component could consist simply of Prague, but homogenized
>>> against appropriately chosen nearby stations in the new EI-corrected
>>> dataset.
>>> Some questions:
>>> 1. Can any of you figure out a significantly better way to construct
>>> a CE-average for the purpose of calibration and combination with
>>> documentary data from CH+DE+CZ? Perhaps some of you would advocate
>>> an unweighted average of all (or selected) individual stations,
>>> rather than constructing four country series and then average?
>>> 2. Reinhard, you know most about all the individual EI-corrected
>>> station records; Are they all of equal quality? Or should some
>>> station(s) be excluded for data quality reasons?
>>> 3. Petr, would it be possible for you to homogenize the Prague
>>> series along the lines suggested above?
>>> 4. We (Petr, Rudolf and I) have had some internal discussions as to
>>> whether we should aim at reconstructing seasonal or monthly
>>> temperatures. Petr advocates seasonal series. I rather think that,
>>> because it is actually possible to construct monthly series back to
>>> 1500, we should produce monthly series for the whole period. We can
>>> of course also produce separately calibrated series for the
>>> traditional seasons (DJF, MAM, JJA, SON) plus the additional JFMA
>>> winter/spring season (for comparison with Stockholm) and also annual
>>> mean series, in order to get the error bars for the seasons being
>>> constructed directly for the seasonal averages. I cannot find any
>>> particular reason why we should not produce monthly series. Even if
>>> they might have poorer quality than the seasonal ones, this will be
>>> accounted for by the error bars. For me, it is not much more work to
>>> do 12 monthly series + 5 seasons + 1 annual series compared to doing
>>> only the seasonal and annual cases. We can of course not discuss all
>>> monthly series in detail in the paper, but we can show them all and
>>> 'deliver' the data.
>>> Looking forwards to hear your viewpoints,
>>> cheers,
>>> Anders
>>> Reinhard Boehm skrev:
>>>> Dear Anders, Rudolf and Petr,
>>>> First some additional information regarding Anders' questions:
>>>> Yes, everything is based on the only existing evidence of a multiyear
>>>> comparison between a preserved original early instrumental site and
>>>> a modern
>>>> one. I would be happy if we could find additional information but
>>>> it seems
>>>> that Kremsm�nster is unique (at least in the GAR I am quite sure
>>>> about it).
>>>> The second thing we did was to develop two other "correction
>>>> models" for NNW
>>>> and N-orientation (the Kremsm�nster EI-site Kremsm�nster is orientated
>>>> towards +30�. Yesterday's new version of the paper-draft tells you
>>>> more
>>>> about that. It is quite a simple "model" but with the existing
>>>> information
>>>> we cannot do better I believe.
>>>> Then we intensively studied metadata and could identify quite a lot
>>>> of the
>>>> three relevant infos we needed for the majority of the EI-sites:
>>>> 1) Height above ground (less important for EI-correction of monthly
>>>> means as
>>>> argued in the paper draft, but interesting perhaps for future
>>>> attempts on
>>>> daily extremes which should be more sensitive to this parameter
>>>> 2) Orientation of the site. This allowed for allocating the single
>>>> sites
>>>> (sunperiods) to one of the three correction models: -5 to +5� was
>>>> defined to
>>>> be N-orientation model, <-5� to NNW and <+5�NNE
>>>> 3) The contemporarily used algorithms for means calculation
>>>> You find the complete collection of metadata and the finally chosen
>>>> correction models on the attached xls-file
>>>> Finally (this is not yet written in the paper draft, but is implied
>>>> in the
>>>> ppt-file through the Regensburg-example) we applied for those cases
>>>> with
>>>> sufficient metadata information the individual correction models
>>>> station by
>>>> station (and sometimes subperiod per subperiod, applied "estimated
>>>> correction models" in unsecure cases (e.g. based on the guess that
>>>> similar
>>>> equipment and habits for means calculation ma be assumed in
>>>> stations managed
>>>> by the same organisation or being in the same country. The
>>>> remaining (few)
>>>> cased with insufficient information were at last adjusted to the
>>>> regional
>>>> mean neighbouring ones with the usual proceeding when homogenising
>>>> without
>>>> metadata.
>>>> Finally what we did not (could not) do was:
>>>> Accounting for different latitudes (within our region I believe
>>>> this to be
>>>> of minor importance, so I would warn to overstreching the
>>>> Kremsm�nster-evidence too much, e.g. to the north (where insolation
>>>> effects
>>>> maybe stronger in summer mornings and evenings) or to the south
>>>> (where the
>>>> astronomical preliminaries tend to cause less problems, but maybe the
>>>> climatological ones (more sunshine in summer) tend to cause more)
>>>> Taking into account heat storage in the measuring walls when
>>>> developing the
>>>> NNW- and th N-correction models. Also for this we did not really
>>>> find the
>>>> necessary information on wall thicknesses, materials, colours
>>>> etc... All in all I think we have done a step into the right
>>>> direction. You will
>>>> learn from the remaining part of the paper (I hope to be able to
>>>> finish
>>>> soon, but there are also other obligations) that the resulting
>>>> EI-corrected
>>>> records fit better to treering-reconstructions, have less
>>>> difficulties to
>>>> explain the high glacier extensions in the Alps in the first part
>>>> of 19th
>>>> century, reduce the previously existing strong difference of annual
>>>> temperature cycle between the 19th and the 20th century (this is
>>>> one of
>>>> Phil's arguments in favour of doing the corrections), it is in better
>>>> agreement with some longterm historic model runs (RIC from the
>>>> GKSS-people
>>>> for example) although the different solar forcing curves (mentioned
>>>> in the
>>>> introduction)are ambiguous (Lean, 200 is in favour of a "cold
>>>> version, Bard
>>>> et al., 2000) tells a story of quite strong solar forcing before
>>>> the decline
>>>> in the 1910s). The high elevation ice cores from the Monte Rosa Region
>>>> (Colle Gnifetti)also provide two different informations, mineral dust
>>>> content tending towards a cooler (EI-corrected) version, stable
>>>> isotopes
>>>> more towards a warmer one near 1800. The existing few infos on
>>>> high-elevation alpine lake sediments tend more to tell a story of warm
>>>> spring and summer temperatzures, but this is the only
>>>> proxy-information
>>>> which really favours the warm solution without corrections. Not to
>>>> forget at
>>>> last the "official" Central European CRU-version published in box
>>>> 3.6, Fig.2
>>>> of WG1-2007 report which shows rather high EI-summer temperatures,
>>>> just as
>>>> our own HISTALP series did before the EI correction.
>>>> SO if you ask me I am convinced that our new version of longterm
>>>> temperature
>>>> records is nearer to the truth than the other existing ones without
>>>> EI-corrections. But I am not sure, whether we should extrapolate
>>>> too much to
>>>> other regions. Series like Berlin or perhaps Paris should be the
>>>> outermost
>>>> locations for it, Southern Sweden may need stronger corrections (as
>>>> you have
>>>> already shown Anders), oceanic locations in UK maybe weaker ones
>>>> and about
>>>> the south I am not sure which factor really dominates, the described
>>>> astronomical or the climatological influences. So I guess for the
>>>> time being
>>>> we have a good solution for Central Europe, but we will have
>>>> difficulties to
>>>> find quantitative information for other regions as long as the
>>>> "Kremsm�nsters" of the North, the West, the South and the East are
>>>> not yet
>>>> existing. How about starting such comparative monitoring? Some of the
>>>> original EI-buildings should still exist.
>>>> OK, that�s all for the moment, I hope it provided the
>>>> Millennium-group with
>>>> some useful facts for the decisions Anders was writing about.
>>>> Best regards
>>>> Reinhard
>>>> P.S.: I cc. this also to Phil, maybe ha also has some comments
>>>> -----Urspr�ngliche Nachricht-----
>>>> Von: Anders Moberg [] Gesendet:
>>>> Montag, 16. Juni 2008 23:34
>>>> An: Reinhard Boehm
>>>> Cc: 'Brazdil Rudolf'; Petr Dobrovolny
>>>> Betreff: Re: AW: Central European temperature reconstruction
>>>> Dear Reinhard,
>>>> cc. Rudolf and Petr
>>>> Thanks indeed for sending this information and corrected data. They
>>>> come very timely! I need not to say (as you already know it) that I
>>>> consider it to be very important that you have developed these
>>>> corrected series.
>>>> I do understand that you will describe how you did the
>>>> EI-corrections in the paper. But can you tell us in a few sentences
>>>> what it is based on? By looking at your presentation (the
>>>> pdf-file), I assume that your 8-yr parallell observations in
>>>> Kremsmuenster forms the basis for your correction. Is this right?
>>>> But how do you apply this knowledge to the other stations? Does
>>>> your correction model account for the orientation (e.g. NE, NW,
>>>> etc) of the thermometer-wall at each site? Do you account for
>>>> observation hours? Latitude? It would be good to get just some
>>>> hints of what you did.
>>>> I also have some comments that I would like address to Rudolf and
>>>> Petr: Given that this paper by Reinhard et al (as far as I
>>>> understand) will be the paper about long instrumental records in
>>>> the CC special issue, it is important that other related papers use
>>>> the new homogenised series for the sake of consistency. In
>>>> particular, the paper about Central European temperatures should be
>>>> as consistent as possible with the new EI-corrected data. We agreed
>>>> some week ago to use a combination of
>>>> [Kremsmuenster+Basel+Prague+Rapp's German average] to construct the
>>>> instrumental regional average. As Reinhard now has developed
>>>> homogenised versions of both Kremsmuenster and Basel, we should use
>>>> these new versions. I assume that you can also use the new
>>>> EI-corrected series in your homogenisation of Prague. As concerns
>>>> the Rapp series for Germany, the situtation is more tricky. We
>>>> could in principle consider using an average of (some of) the
>>>> German stations in Reinhard's dataset (Hohenpeissenberg, Karlsruhe,
>>>> Munchen, Regensburg, Stuttgart). But then we would miss information
>>>> from more northern parts of Germany. I actually think it is more
>>>> important to have the EI-corrected German series from southen parts
>>>> of Germany than having un-corrected Rapp data for all Germany. We
>>>> might consider some compromise? E.g. homogenising the Rapp series
>>>> with the EI-corrected German series as reference series. Another
>>>> thing; we could also consider using more station records from
>>>> Switzerland and Austria than only Basel and Kremsmuenster. I think
>>>> we should have a little email conversation about this, so that all
>>>> of us can say our opinions before we decide on how to construct an
>>>> appropriate Central European temperature series to be used for
>>>> calibration of the documentary data from CZ+DE+CH. What is your
>>>> opinion -I mean all three of you?
>>>> Regards,
>>>> Anders
>>>> Reinhard Boehm skrev:
>>>>> Dear Anders, dear Rudolf
>>>>> Please understand my delay in answering your questions and your data
>>>>> request. The reason war quite a complicated decision finding process
>>>> mainly
>>>>> with the Italian co-authors about our definite result in
>>>>> correcting the
>>>>> EI-bias in early warm-season temperatures.
>>>>> But last week we came to the decision about our definite new
>>>>> version of 32
>>>>> Central European temperature series starting before 1850. And I,
>>>>> as the lead-author finally decided to write the paper as part of
>>>> your
>>>>> planned Millennium-publication, Rudolf asked me to contribute to
>>>>> some time
>>>>> ago.
>>>>> So I have finally started writing and for Your information I
>>>>> attach the
>>>>> recent (yesterday's) status of the paper plus a presentation I
>>>>> gave about
>>>> it
>>>>> (see at pages 7 to 12 of the attached Mondsee-pdf).
>>>>> I also attach 2 xls-files. The first contains all 32 EI-series in the
>>>> region
>>>>> in three different modes: (from left to right) EI-corrected,
>>>>> homogenised
>>>> but
>>>>> not EI-corrected and the original series (only outlier corrected)
>>>>> The second xls-file explains the station code and it also tells
>>>>> you whom
>>>> you
>>>>> should acknowledge as data provider, if you explicitly use one of
>>>>> these
>>>>> series in the publication.
>>>>> Best regards
>>>>> Reinhard
>>>>> -----Urspr�ngliche Nachricht-----
>>>>> Von: Anders Moberg [] Gesendet:
>>>>> Mittwoch, 28. Mai 2008 11:38
>>>>> An:; Dr. Ingeborg Auer
>>>>> Cc: Brazdil Rudolf; Petr Dobrovolny; Rob Wilson
>>>>> Betreff: Central European temperature reconstruction
>>>>> Dear Reinhard and Ingeborg,
>>>>> cc. Rudolf, Petr, Rob,
>>>>> Hope all is well with you! As you probably already know, I am
>>>>> working together with Rudolf, Petr, Rob and other colleagues in
>>>>> Millennium with writing papers for the Special Issue in Climatic
>>>>> Change that Rudolf is organizing. I have understood that Reinhard
>>>>> is contributing with a chapter about long instrumental records.
>>>>> One of the chapters will deal with reconstruction of monthly
>>>>> temperatures for Central Europe back to 1500, based on index
>>>>> series derived from documentary evidence from Germany, Switzerland
>>>>> and the Czech Republic. In the last few weeks, I have been engaged
>>>>> in the problem of calibrating these data against long instrumental
>>>>> data for Central Europe, and to put error bars on them. The basic
>>>>> principle for the calibration is simple linear regression for a
>>>>> period of overlapping data. The attached figure shows you some
>>>>> preliminary results for seasonal averages (month 13 = DJF, 14=MAM,
>>>>> 15=JJA, 16=SON). The error bars are 2*sigma error from
>>>>> calibration, with additional adjustment for changes in running
>>>>> correlations between the individual country series. (The error bar
>>>>> estimation for these data is far from trivial, but this is not
>>>>> what this email is about ...)
>>>>> So far, the instrumental data I used are taken from the CRUTEM2v
>>>>> dataset, with an extension back to 1781. Phil Jones and I created
>>>>> a long record for Central Europe (CE) from these data, for a paper
>>>>> we published in 2003. I use the same series here. However, we are
>>>>> discussing if we should rather use another average temperature
>>>>> series for CE which we develop specifially for this paper. This is
>>>>> the real reason for writing to you: We would need the best
>>>>> available long instrumental records from Germany, Switzerland and
>>>>> the Czech Republic to construct a suitable CE series. Would you be
>>>>> interested in contributing with a selection of series from
>>>>> HISTALP? Data from Austria can also be considered, even if no
>>>>> documentary data from your country are used, because Austria fits
>>>>> well into the climatic region of interest.
>>>>> We are of course aware of the problem of possible too warm summer
>>>>> temperatures before mid-19th century. As you can see in the
>>>>> attached plot, the reconstructed (smoothed) JJA temperatures lie
>>>>> consistently above the zero line (1961-90 average). Obviously, if
>>>>> the instrumental temperatures are biased in the calibration period
>>>>> (1781-1820), then the entire reconstruction will also be biased.
>>>>> If you have any instrumental series in the pipeline that are
>>>>> corrected for this bias, then it would be excellent if we could
>>>>> use them. Maybe you are planning to present such series in your
>>>>> chapter of the Special Issue?? If so, it would be a very nice
>>>>> connection between the two papers, if we can use the same data.
>>>>> If you don't have any such corrected data, then we can at least
>>>>> point out the problem and discuss it properly. In any case, we
>>>>> feel that it would be good to build a new instrumental CE
>>>>> temperature series, to be used here rather than the extended
>>>>> CRUTEM series. Many early data records in the CRUTEM dataset have
>>>>> not been subject to homogeneity testing. Rather, the dataset is
>>>>> just a collection of what Phil has been able to collect. (For
>>>>> example, it contains Austrian series from ALOCLIM and data from
>>>>> IMPROVE, but there are also other more or less untested early data).
>>>>> We hope that you would like to collaborate by contributing with
>>>>> your most appropriately selected station records. If you would
>>>>> like to do this, then we would need the data very soon as we are
>>>>> approaching the deadline for the paper and other chapters in the
>>>>> special issue are dependent on using our reconstruction. So,
>>>>> please, answer as soon as you can and tell us if, and how, you can
>>>>> contribute. I am convinced that Petr (who is the lead author) is
>>>>> happy to include you as co-author(s) if you like.
>>>>> Looking forward to hear from you soon,
>>>>> Best regards,
>>>>> Anders


No comments:

Post a Comment