date: Tue Nov 30 16:24:32 2004
from: Phil Jones <p.jonesatXYZxyz.ac.uk>
subject: Re: Section 3.5
to: Kevin Trenberth <trenbertatXYZxyz.ucar.edu>, b.sodenatXYZxyzmi.edu
The dimming box needs to be consistent with what is in Dave Easterling's
section 3.3.4. I've not got to that yet - will do tomorrow.
Will look at this email and others tonight.
At 16:01 30/11/2004, Kevin Trenberth wrote:
Thanks for responding, some follow up discussion follows
Brian Soden wrote:
A quick response and some specific comments on your conecerns below.
I agree that the total cloud cover trends from surface and ISCCP are not consistent with
each other, and believe that this discrepancy should be made clear and suggest
including a figure just for this purpose (ie, supplementing Aiguo's original version of
I also agree that ISCCP has known problems, however I do not believe that we know how
these problems affect the trends (more on this below). It has been shown that ERBS
radiation trends are reasonably consistent with those computed from ISCCP data - this
may be fortiutous, but again I don't believe that it should be dismissed simply based on
speculation. There needs to be published/submitted results to base it on. I don't
believe that Garret Campbell's work provides a definitive answer to this (at least I
wasn't covninced by what I saw in Paris) and apparently Garret does NOT want the results
of his submitted Science paper cited in the ZOD (at least according to Steve Warren) -
not sure why. There is also the increase in surface solar radiation data after 1990
which is argued to be qualitatively consistent with the decrease in ISCCP cloud cover
since the late 1980s.
I am not very familiar with ISCCP and all the geostationary satellites. My
understanding is that there have been instrument upgrades and the frequency response
has varied considerably from satellite to satellite. It means, if true, that trends are
very difficult to do. That is not speculation. Garret has published the article below
in the grey literature:
Campbell, G.G., 2004: View angle dependence of cloudiness and the trend in ISCCP
cloudiness (extended abstract). Thirteenth Conference on Satellite Meteorology and
Oceanography, American Meteorological Society, Boston, P 6.7. (Available from
I am not sure this is the whole answer at all, it is more the changes in instruments and
their degradation with time that concerns me.
I am not sure about the increase in sfc solar: here you are referring to the
Ohmura/Wild stuff? That is at fairly few stations. Please also note the box I prepared
some time ago: you should have a copy on dimming and the evaporation vs potential
evaporation (pan): I'll attach it. The box needs to be consistent with the main text.
We obviously need to better understand the extent of agreement/disagreement between the
ISCCP and surface clouds. I believe Joels work in showing agreement between ISCCP and
surface high clouds cover trends is important and obviously relevant here. It is
presumably the low clouds where ISCCP and the surface observations disagree. Low clouds
are also the most likely culprit for the ERBS radiation budget trends (which are
dominated by SW more than LW). Perhaps we could create maps of the trends in total,
low, and high cloud cover for ISCCP and surface clouds (or maybe Joel has already done
I agree that there may be differences in low and high clouds. The HIRS and ISCCP
disagreements hinge on thin cirrus detection, don't they? ISCCP had major problems with
aerosol from Pinatubo and there is a trade off in ISCCP between aerosol and cloud. The
spurious changes in cloud in ISCCP because of the aerosol are well known (though I don't
know the reference). Changes in high cloud: cirrus without changing total amount can
indeed affect OLR.
More comments below:
I also do not believe Wielicki's results. It was suggested in my published comment in
Science on this, but I did not get it quite right. It turns out almost the entire trend
in OLR is because of a jump in the ERBS record in late 1992. That jump occurs relative
to AVHRR and HIRS. Granted there are major problems in the long-term record of both of
these - although AVHRR has been reprocessed by Jacobowitz et al 2003 (which is not
cited) - but both were stable and on a single satellite at the time of the jump. To
then say the trends from Wielicki et al are right by using ISCCP just doesn't stand up
according to my reasoning.
The largest trend in the ERBS edition 2 (altitude corrected) radiation data occurs in
the SW. The LW trend/jump is quite modest (barely outside the model envelope). One
could argue that there is a jump in the SW in late 1992, but it is complicated by the
large Pinatubo anomalies of the opposite sign immediately preceeding it.
Please see the attached figures in ppt. The first one is the one showing the ERBS vs
HIRS and AVHRR with the time of the jump highlighted by the arrow. This is the original
processing of the ERBS by Wielicki et al. There are spurious trends and low frequency
variability in the blue and green curves from changes in satellites and instruments and
lack of calibration, but at the time of the arrow it was stable. This highlights the
jump. The second slide shows the original and edition 2 version. It is important to
note that only 2 out of the 3 curves are independent: the Net = OLR+ ASW. In these
versions the net is constant: no change. But equal and opposite jumps in SW and OLR (in
contrast to your assertion). In edition 3 there is now a drift in net owing to the
altitude correction??? and the jump is not quite as apparent. In the third panel is a
figure provided by Wielicki on the calibration (red curve) that has been subtracted from
the raw OLR signal to produce the blue curve, which is the published OLR record.
In my original comment, when I did not have a complete perspective, I focussed on the
break in 1993 when the battery failed and the instrument did not work (the blue line is
a linear interpolation). It took 3 months to get the new battery in place and the
system working. Meanwhile the space craft cooled off and the hot point temperature went
down (different rotation of satellite as it orbits) so hence the jump in calibration
then. But note also the spike in late 1992 at the time of the jump in Fig 1 (pink
I did not include the reprocessed AVHRR data from PATMOS (Stowe et al 2002, Jacobowitz
et al 2002) because it is explicitly detrended to remove the effects of equatorial
crossing time drift (see attached figure). This obviously removes any long-term trends
as well, so it seemed a bit silly to note that a detrended cloud product had no trends.
The equatorial crossing time drift aliases an artifical trend in cloud cover of ~0.5 to
1%/yr in polar orbiter data. This should also affect both Garrett Campbell's AVHRR-only
version of ISCCP and Wylie's HIRS product. In fact, it may explain why Wylie's data
shows an increase in cirrus, while the surface and ISCCP data suggest a decrease .... so
the impact of equatorial crossing time drift demonstrated by Stowe et al probably
should be cited.
No this is not correct. The PATMOS reprocessing did use some vicarious calibration
(Libyan desert), and also recognized the 2% cloud change per hour drift in ECT. It used
this regression to remove the spurious trends associated with the ECT drift. It did NOT
remove trends otherwise. Interannual and longer term variability is not otherwise
suppressed. The time series is a legitimate alternative view of the cloud cover change
with time that disgarees with ISCCP from 20N to 20S. Their OLR also disgrees with the
ERBS, although there are obvious problems with their OLR for NOAA 9.
It is worth highlighting the remarkable sensitivity of the cloud cover the the exact
time of day of observation: the diurnal cycle is remarkably large. And this discrepancy
is another of the major disagreements with ISCCP.
In the writeup there is circular reasoning. The radiation is used to argue about the
clouds and the clouds are used to argue about the radiation. I would prefer to see the
clouds self contained and deal with clouds, and the synthesis to take place later.
I feel that it is inefficient to discuss clouds exclusively in section 3.4.3 and
radiation exclusively in section 3.4.4. It seems natural to mention the ERBS trends
when discussing the ISCCP trends and vice-versa. Likewise, it seems natural to mention
the consistency with precipitation / DTR when discussing the surface cloud trends, and
to mention the consistency with surface clouds when discussing the surface radiation
trends. The agreement with indpendent correlative data is important when assessing the
credibility of the trend. I've tried to treat consistency between ISCCP and ERBS in
the same manner as I treated the consistency between surface cloud/DTR and surface
This is true if there were clear signs as to what the changes are. But there are not:
both the clouds and the radiation are embroiled in uncertainty. One could just as
easily cite the disagreement with the sfc obs. So it seems like there is a need to
discuss each on their separate merits and then try to pull them together and
synthesize. That is indeed where the two help to bolster the argument that something is
You in fact note that AVHRR and HIRS cloud trends to not agree with ISCCP. But special
scrutiny is warranted at the time of the jump, which is published online as part of the
Wielicki article actually. I can send that figure if you like.
I agree, but the jump in LW is actually very small in the altitude corrected vesion of
ERBS. One would now want to focus on the SW, however neither HIRS or AVHRR have SW
See figures attached. The evolution of the product matters here.
I am also uncomfortable about Norris work. You cite 2 of his publications that are not
yet submitted or under review. I have not seen them. I though we should have him as a
CA as he has done a lot of work in this area. But I have to say I do not fully trust
his results or analysis. This distrust comes from working with him directly as a
post-doc at NCAR.
I am also uncomfortable with listing "to be submitted" work, but I was very clear with
Joel on the requirements for listing the references so obviously he feels it will be
submitted in time. I do feel his work on comparing ISCCP and surface observations is
very important, since that seems to be a key area of discpreancy. As an aside, when
Joel was at GFDL he was very skeptical of the ERBS trends for just the reasons you
cited. He was also the first to identify some of the key artifcacts in the ISCCP trends
(e.g., the strong dependence of the trends on viewing angle). He pointed this out to
Cess and Udelhofen (when it was under review) and they then went back and recomputed the
ISCCP trends with the regions of large zenith angles masked out (and still found a
reduction). So I believe Joel is naturally skeptical of observational data sets in
general and analyzes them carefully.
I presume you have copies of his unpublished stuff? I do not.
I agree the figure and discussion of the radiation vs ocean heat content goes a bit
outside our territory, but I do believe that the consistency of ocean heat content and
ERBS net radiation should be mentioned.
I agree and we need to make that happen but probably not in our chapter??? Unless it is
in section 3.9. Will require interactions with chapter on oceans.
I'd welcome prespectives from others cc'd on this also.
Kevin E. Trenberth e-mail: trenbertatXYZxyzr.edu
Climate Analysis Section, NCAR www.cgd.ucar.edu/cas/
P. O. Box 3000, (303) 497 1318
Boulder, CO 80307 (303) 497 1333 (fax)
Street address: 1850 Table Mesa Drive, Boulder, CO 80303
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