cc: "Willett, Kate" <kate.willettatXYZxyzoffice.gov.uk>, Peter Thorne <peter.w.thorneatXYZxyzglemail.com>
date: Tue Jun 23 12:18:29 2009
from: Phil Jones <p.jonesatXYZxyz.ac.uk>
subject: Re: [Fwd: 2009JD012442 (Editor - Steve Ghan): Decision Letter]
to: email@example.com, Dick Dee <Dick.DeeatXYZxyzwf.int>
Emails to Kate yesterday were returned by the ECMWF server (for your email
address) but not for Dick's?
I also found the two emails you sent last night in my spam list. No idea why
this is happening. I found some other semi-important emails in my spam as well!
Anyway - hope you get this email!
All three reviewers are positive, which is good, but there is still a lot of to do as
Here are some initial thoughts. Before I begin - it seems as though Rev 2 comments have
ended abruptly during #13. I'd suggest you ask if there is any more?
I would have thought that the second point (larger trends in full ERA-INTERIM fields) was
just an interesting aside, and not as important as the RH decline.
I'll need to go back to see if sections 5 and 6 can be reordered/restructured?
Both Reviewers 1 and 2 (they appear to be Kevin and Aiguo, but odd to have two people
who only live a few rooms apart!) make quite a few statements about GPCC. We're
doing updating work on the higher resolution CRU-TS (0.5 by 0.5 degree lat/long)
datasets. We're doing comparisons with GPCC and for the Giorgi type regions (as in
Fig 3.14 of Ch 3 of AR4) and the agreement is amazingly good. Maybe all you
need to point to is this Figure and the previous one (Fig 3.12) to say that for land
regions at the continental scale, it doesn't matter which datasets are used (for the
from the 1970s). The key thing is that they just use gauges, with no satellites.
My view is that bringing in satellites as in CMAP and GPCP products can lead to
problems, and some circularity with ERA results - as you'll be using some of the same
satellite data products. The point to emphasize for precip is that GPCC is totally
independent from any ERA (40 or Interim) input.
I've come across these issues about GPCC before. I've been haranguing Bruno Rudolf
and now Tobias Fuchs of GPCC to write something up for a number of years within AOPC!
I think their QC is likely the best of all the centres, but they will continue to get
doubts if they don't write anything up. They should at least explain how they do their
interpolation - it can certainly be done better.
GPCC is using so much more data that is has to be better than any other product.
They can't release the raw station data, and it seems they can't release the numbers
in each grid box.
There will be an HC paper on the buoy/ship SST issue, but this isn't yet used
It will come, but not before your paper goes back.
I hope it is fairly straightforward to do RMSs as well as correlations. We had SDs in the
2004 paper. I don't think RMSs would show anything untoward, but would take up some
WRT Rev 2, I'm not that convinced by some of Aiguo's arguments. Between us, I'm
not that convinced by some of his data analyses. The ones involving PDSI leave a
lot to be desired (this is coming to light in other work we are doing).
Rev 2 #6 Obviously not read the paper(s). CRUTEM3 is a simple average of stations
within a grid box. There is no interpolation! If there are no stations, then there is no
I think this is the same for HadCRUH as well.
Rev 2 #13 Comment seems to end abruptly. I'd like to know what I might
have said! I don't think I've ever said I doubt GPCP!
I am around all the time except for the week of July 12-17, when I'll
be at the IPCC Scoping meeting in Venice. Kevin will be there as well.
Aiguo will be in CRU the first few days of the week after (July 20/21)
At 22:53 22/06/2009, Adrian Simmons wrote:
It's a bit irritating getting a review one wants to nail just before leaving for
Brussels for three days of EC-related meetings.
I'm sure now that reviewer 2's comments on SYNOP numbers is easily answered. The number
of GTS SYNOPs went up a lot, but that's not because there were a lot more stations
installed - the existing one just started having their data transmitted more frequently
than 6-hourly. But this should hardly have effected the RH2m analysis as it uses only
the 0, 6 , 12 and 18UTC obs that have been there pretty well all the time. It only uses
off-time obs if the value for the main synoptic hour is missing. The 4D-Var does
assimilate more data over time, but here we appeal to fig 8 and argue that the increment
does not shift over time. We already argue in the Appendix that the extra obs over North
America may well be part of the difficulty HadCRUHext has for that region.
Anyway I'd like to confirm that the number of used SYNOPs does not change much over time
for the OI RH2m analysis. I know how to find the number in the job output, but I don't
know how to retrieve the job output from the logfiles stored in ECFS. I would only look
at a few samples. I'd be grateful if you'd let me know how to do this.
In any case even if there was a problem with the numbers increasing sharply around 2000,
this would manifest itself in a sudden drop in the RH time series, not a steady decline
over the last few years.
After a bit of thinking I can find several things wrong with reviewer 2's argument why q
over land is insensitive to variations in q over sea (think coastal mountain ranges,
deserts, drought regions - moisture does not simply build up everywhere over land via
onshore winds from the boundary-layer until it rains), and the response can draw
attention to other points made in the paper, such as the coherence of changes in the
vertical, and the similarity (but lag) of the q series over land and sea. Hard to
believe the latter is all coincidence.
Also, there is a relationship between q and precip, not generally strong, but there's a
high correlation for Australia.
Better stop for now.
-------- Original Message --------
Subject: 2009JD012442 (Editor - Steve Ghan): Decision Letter
Date: Mon, 22 Jun 2009 16:42:51 UT
Manuscript Number: 2009JD012442
Manuscript Title: Low-frequency variations in surface atmospheric humidity, temperature
and precipitation: Inferences from reanalyses and monthly gridded observational datasets
Dear Dr. Simmons:
Attached below please find 3 reviews on your above-referenced paper. One of the
Reviewers has raised questions and made suggestions for important revisions, mostly
involving organization and presentation. Please consider the Reviewer reports carefully,
make the necessary changes in your manuscript and respond to me, explaining how you have
addressed these comments. In your Response to Reviewer letter, please include a
statement confirming that all authors listed on the manuscript concur with submission in
its revised form.
The due date for your revised paper is July 20, 2009. If you will be unable to submit a
revised manuscript by July 20, 2009, please notify my office and arrange for an
extension (maximum two weeks). If we do not hear from you by the revision due date, your
manuscript will be considered as withdrawn.
When you are ready to submit your revision, please use the link below.
*The link below will begin the resubmission of your manuscript, please Do Not click on
the link until you are ready to upload your revised files. Any partial submission that
sits for 3 days without files will be deleted.
(NOTE: The link above automatically submits your login name and password. If you wish to
share this link with co-authors or colleagues, please be aware that they will have
access to your entire account for this journal.)
**In order to save time upon acceptance, it would be helpful if files in the correct
format are uploaded at revision. Article and table files may be in Word, WordPerfect or
LaTeX and figure files should be separately uploaded as .eps, .tif or pdf files. If you
have color figures, please go to the site below to select a color option. Please put
your color option in the cover letter.
Editor, Journal of Geophysical Research - Atmospheres
-----------Important JGR-Atmospheres Information------------------------------
Submission, Review and Publication Stages Chart
Text Preparation and Formatting
Acceptable Electronic File Formats
Editorial Style Guide for Authors
Auxiliary Materials (Electronic Supplements)
Guidelines for Preparing Graphics Files
Prices for Color in AGU Journals
AGU Copyright Transfer Form
Manuscript Status Tool (for manuscripts recently accepted)
If you need assistance with file formats and/or color charges please e-mail
jgr-atmospheresatXYZxyz.org and quote your manuscript number.
If you need Adobe Acrobat Reader to download the forms, it is available, free, on the
internet at: http://www.adobe.com/prodindex/acrobat/readstep.html
Reviewer #2 (Comments):
Review of JGR Manuscript entitled
Low-frequency variations in surface atmospheric humidity, temperature and precipitation:
inferences from reanalyses and monthly gridded observational data sets
by A.J. Simmons, K.M. Willett, P.D. Jones, P.W. Thorne, and D. Dee
This paper provides a nice and useful summary on how the ERA-40 and ERA-Interim surface
analysis products of temperature and humidity were derived, and a fairly comprehensive
evaluation/comparison with the HadCRU surface data sets derived purely from surface
observations, as well as with three other precipitation products. They found that in
general the ERA surface temperature and humidity data from 1973 onward are in close
agreement with the HadCRU data sets and that ERA precipitation also follows closely with
gauge-based products, although long-term changes differ. Furthermore, the study reports
a significant and steady decline in surface relative humidity (RH) over land from
~1999-2008 and suggested that the recent steady SSTs might be responsible for this land
RH decrease. The manuscript is well written, the analysis appears to be comprehensive,
and the results are of interest to many readers in the climate community. I think the
paper should be published after some relat
My main concern is the interpretation of the recent RH decline over land. To me, the RH
decreases shown in Fig. 4 and Fig. 7 look a bit spurious (non-climatic, e.g., lack of
variations in Fig. 4 and stepwise changes in Fig.7) rather than realistic changes. They
are also inconsistent with the RH changes during recent decades (up to 2004) reported in
Dai (2006, JC), and this is not pointed out in the paper. As shown in Dai (2006), there
was a 3-fold increase around the late 1990s in the number of surface humidity reports
(mostly in North America but also over some other regions) included in the WMO SYNOP GTS
reports. Furthermore, I personally found that there were other (undocumented) changes in
the SYNOP reports around that time that led to shifts in derived precipitation and cloud
frequencies over Euroasia and other places. Thus, there are reasons to suspect some
non-climatic changes in the SYNOP reports around the late 1990s that might alter the RH
trend over land.
I also was not convinced by the physical explanation of the RH decline (p.23). Even if
the surface q stayed the same over the oceans during the 1999-2008 period when land air
temperature has been increasing, this can not explain the RH decrease over land. This is
because as long as the marine air contains more water vapor than continental surface air
(which is still true even if marine sfc. q did not increase), advection of marine air
onto land should cause land q to accumulate and RH to increase until the land q and RH
reach certain levels so that precipitation kicks in to remove the moisture over land.
Remember that the atmospheric moisture storage (PW) is very small compared with the
annual P and E fluxes, thus any perturbation in RH is quickly (within days) restored
through surface E, vertical mixing, or lateral advection/mixing. If the RH in the marine
air had decreased, then land RH would likely to decease too. Dai (2006) did not show RH
decreases over oceans since the
1980s. I wish the authors of this paper would also show RH series over ocean, at least
since the middle 1980s.
For the ERA humidity data, the large well-known inhomogeneities in radiosonde humidity
records will certainly propagate into the ERA background forecast and its analysis
fields, making them not really suitable for long-term trend analyses. For example, all
U.S.-operated radiosonde records (including many in the Pacific) before about Oct. 1993
report a dew point depression (DPD) of 30deg.C or a RH of 20% for any cases where RH is
below 20%, which resulted in an abnormally higher frequency of reports of DPD=30deg.C
and few reports below and no reports above DPD=30deg.C. This practice is also found in
some Mexican, Canadian, Australian, and few other places (but stopped at different times
from the late 1980s to the 1990s). In general, the newer humidity sensors during the
last 10-15 years report more low RH or large DPD cases, whereas earlier ones had no
measurements or incorrect values for these cases. One can see this shift in the
histograms of daily DPD made by different humi
sensors. Thus, one needs to be very cautious when radiosonde humidity data are used in
assessing trends, even if they are used indirectly (as in the ERA surface humidity
Some other comments:
1. Abstract: it gives the impression that even the long-term mean values for surface T,
q and RH are the same between ERA and HadCRU data sets, which appears to be not the case
as the respective means are removed in all plots. Please mention that the climatological
mean may differ (if this is the case) even though the anomaly variations are similar.
2. Abstract, at the end: Please note that the mean precipitation amount and its change
rate are not controlled by atmospheric water vapor amount (q), although higher q is
often associated with higher P (e.g., tropical vs. high latitudes). Locally, you can
have moist air passing by without any rain. Globally, annual P is controlled by how much
moisture gets evaporated from ocean and land surfaces (i.e., P=E), and this surface E is
primarily controlled by surface energy terms. In essence, P and E are water fluxes, and
PW (or q) is the water storage in the atmosphere. People often link P to q because of
the associated mentioned above (through low-level moisture convergence in a storm,
etc.), and think that P change rates somehow should follow that of q or PW. However, and
P (or E) and q are controlled by different processes and in general the flux terms are
not coupled with the storage terms in a cycling system (e.g., no one would think P or E
is controlled by water storage in t
3. p. 3, top: the net radiative effect of clouds is relatively small, when their effect
on solar radiation is included. To include clouds in the natural greenhouse warmth is a
bit misleading because the higher surface temperature is maintained primarily by the
greenhouse effect of water vapor and CO2.
4. p. 4, middle: Again, any sampling/reporting biases in WMO SYNOP reports could affect
both ERA and HadCRUH humidity data. Thus caution is still needed.
5. pp.5-6, section 2a: So in essence, ERA-40 and ERA-Interim surface T, q, and RH are
another analysis product based on surface observations, just like the HadCRU and other
climate data sets. The only difference is in the analysis methodology (IO interpolation
with the use of the ERA background forecast fields vs. other more conventional analysis
methods). Like most users, I thought the ERA surface fields are more tightly coupled
with the reanalysis model system. I think it would be helpful to point out the above at
the beginning of this section or in the Introduction.
6. p. 7, top: Please briefly mention how the station anomalies were aggregated onto
5deg. grid in CRUTEMP3, e.g., by simply averaging station values within the grid box, or
making use of correlated, nearby station data outside the box when sampling inside the
box is sparse? I think most people would use the later to increase the coverage in the
7. p. 7, bottom: Have any adjustments/corrections done for the most recent decades
(1999-2008) in HadCRUH+ext? This is the period when RH decreases. Are there any
homogeneity issues in combining the extended records with the homogenized HadCRUH?
8. p. 9, top: How could the fit of the ERA background forecasts capture multiple shifts
induced by instrumental changes or reporting practices, especially when the future
changes are needed to determine the timing and the size of a shift. Many statistical
methods specifically designed to do these two tasks by analyzing the whole historical
series still have difficulties in reliably detecting the locations of shifts and can
only make a best guess regarding the real shift size. I wonder how one can do this in a
reanalysis system when future records are not used yet, or nearby station series are
combined together to form a grid box series that contain shifts from multiple stations
(i.e., the stepwise patterns become very complex and look more like real variations).
9. p. 9, middle: I can't believe the GPCC people are still gridding precipitation total,
not anomalies. This makes their products useless for long-term change analyses. Another
land precipitation product from 1948-present that is derived from gauge records and the
OI method is the PRECL from the NCEP Climate Prediction Center (CPC, ref: Chen et al.
2002, J. Hydrometorol.). I think that is a better products for assessing long-term
changes in land precipitation, although the gauge coverage for recent years (after 1997)
may be not as good as that of the GPCC.
10. p. 11, middle and bottom: need to point out in Abstract or Summary that differences
in the mean exist between the ERA and HadCRU T and humidity data.
11. Fig. 1 and other Figures: I suspect that different mean values were removed in
computing the difference series. If that's the case, then need to point out this (i.e.,
the difference is between the anomalies relative to their respective mean).
12. Fig. 4: also show RH over the oceans for the last 25 years?
13. Fig. 11: with the changing gauge coverage and gridding precipitation total, one can
not trust the low-frequency variations in the GPCC products. Phil Jones and other have
Reviewer #3 (Comments):
Review of the paper entitled "Low-frequency variations in surface atmospheric humidity,
temperature and precipitation: Inferences from reanalyses and monthly gridded
observational dataset" by A.J. Simmons, K. M. Willett, P. D. Thorne and D. Dee.
Recommendation: Accept with minor changes.
Summary of the paper:
This is an elaborate study examining trends in temperature, humidity and precipitation
from the latest ECMWF reanalysis, comparing with independent gridded analyses, which are
also performed with utmost care. The paper revealed that the commonly accepted
assumption that the relative humidity stays the same under global warming condition does
not necessarily holds over land. This is an important finding and should be of interest
to wide climate communities. There are several other important contributions, such as
the sensitivity of observation coverage on long term trend, which can only be studied by
the use of reanalysis that has full global coverage. This paper also presents that the
ERA-40 and ERA-Interim are of very high quality and useable for low frequency climate
1. I am particularly impressed with the way the work is performed. This is a very
elaborate work using a variety of datasets to present that there is a strong long time
trend in temperature and humidity. This thorough work made it possible to convince
readers these observed facts. Although the finding of the decrease in relative humidity
over land is credible, it may be more meteorologically interesting and convincing if
additional analysis is made to present the possible mechanisms of the absence of
increase in specific humidity over land. If reanalysis is used, it is not impossible to
estimate the change in the moisture transport into land areas (although this may involve
considerable amount of work). It may also possible to examine the change in large scale
mean land-ocean circulation that contributes to the transport of moisture. From
heuristic point of view, stronger heating over land tends to strengthen upper level high
and subsidence, which may prevent moisture to be
transported inland, and such trend may be detectable from large scale reanalysis. In
terms of the change in precipitation, moisture availability and relative humidity are
important, but static stability and large scale convergence should also play an
important role. If any of these additional analyses can be performed, or even discussed
in qualitative manner, it will enhance the paper.
2. It is not very clear how the diurnal variations of temperature and humidity are
handled in this study. It is helpful to state the time frequency of reanalysis output
that is used to compute daily mean, and the way observed daily mean are obtained.
3. Are there any reason that the relative humidity or dew point depression is analyzed
and not the specific humidity itself?
4. The paper is a little too long. One way to shorten it is to separating it into two
parts by adding analysis suggested above, or separating the analysis of precipitation.
This is just a suggestion and decision is up to the authors.
1. Page 6 & 11. The authors claim that the use of anomaly will reduce the influence of
surface elevation differences. Can this be true even the relation between elevation and
relative humidity/specific humidity is very nonlinear?
2. It may be friendlier to the reader why relative humidity and specific humidity are
both examined. Some introductory remarks on the different impact of relative and
specific humidity will help.
3. Page 13. Lines 298-300. These lines just present why the ERA-40 and Interim are
different but not the reason for the ERA-Interim worse than ERA-40 over Africa.
4. Page 14. Lines 316-328. Is it possible to separate the actual reduction in the number
of observations and the reduction in data used by CRUTEM?
5. Page 15. Line 364. It seems that the difference in analysis between ERA-40 and
ERA-Interim seems to be used as a measure of the reanalysis accuracy. Is this a good
6. Page 17. Lines 392-397. Can it be possible to mathematically estimate the relation
between the correlation of specific humidity and relative humidity? Since relative
humidity is a function of specific humidity, temperature and pressure, it seems natural
that the correlation for relative humidity should be lower. However, this will depend on
which parameters are analyzed in the first place.
European Centre for Medium-Range Weather Forecasts
Shinfield Park, Reading, RG2 9AX, UK
Phone: +44 118 949 9700
Fax: +44 118 986 9450
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.jonesatXYZxyz.ac.uk