cc: Aad van Ulden <uldenvan@knmi.nl>, "Klein Tank, Albert" <Albert.Klein.Tank@knmi.nl>
date: Fri, 19 Jan 2007 18:34:41 +0100
from: Geert Jan van Oldenborgh <oldenborgh@knmi.nl>
subject: Re: Precipitation trends statement IPCC 4AR SPM
to: Phil Jones <p.jones@uea.ac.uk>, David.Easterling@noaa.gov

<x-flowed>
Dear Phil, David,

thank you you for the figure, which clarifies a lot.

Phil Jones wrote:
>      If you've just seen the SPM, then you will not know about a figure
>  within Chapter 3. These are figures 3.14 and 3.15. I'm not supposed to
>  send these out, so you got them from Albert. Don't pass on to
>  anyone else.
> 
>     So the SPM bullet points are based on these. There are also trend
>  maps by seasons for 1979-2005 and the year for 1901-2005 and 1979-2005,
>  and global land series time series for 1901-2005 from various databases 
> - many more than just GHCN and CRU.

However, the reader of the SPM will not know these maps either, and 
assume something else from the names than you mean.  Also you do not 
follow this consistently: "Western Africa" with a clear trend is 
arbitrarily replaced by the Sahel, with only a trend in the western 
half, and "Southern South America" is replaced by "eastern SOuth America".

>    We chose the regions in the chapter to show precip differently from 
> how it
>  had been done in previous IPCC reports. The regions were defined in a 
> paper
>  by Giorgi and someone else (from about  2001/2002) that is used in Ch 11
>  (Table 11.1).

Why are the regions not defined based on the signal?  This way one 
groups together regions with and without trends (e.d., in the 
Mediterranean, with no significant trends on the European side).

>    The regions are large, take no account of rainy seasons or rainfall 
> regimes,
>  so they have very little climatological content. They use a lot of the 
> gridded
>  data though and there are some surprising similarities and dissimilarities
>  between them. We chose annual, as we only had space for one Figure.

This seems like a wise choice, especially since some of the observed 
changes are shifts in the rain seasons, as in the southern African rain 
season moving backwards.  A fixed window like JJA would show a decrease 
when in fact there is none.

>    Now the important point - the SPM. If you've read the SPM you'll have
>  noticed that hardly any country is mentioned. This is deliberate and we
>  refer to large regions. This is because we would likely not get the 
> text past
>  the govts in Paris the week after next if we were that specific.

I agree that this makes sense, however, I disagree with choosing the 
regions first and making statements as if the observed trend applies to 
the whole region, rather than parts of it.  The reader of only the SPM 
will conclude that rainfall has decreased everywhere in the 
Mediterranean, when in fact it has not in half; same with Central Asia, 
Eastern North America, etc.

Coming back to some individual regions mentioned in the SPM I still do 
not understand most of the claims made in the SPM statement.

1) Eastern North America: your figure shows as well as my maps that a 
significant increase is only seen in the easternmost provinces of 
Canada.  This should not be labelled "Eastern North America"; there is 
no trend in New York City and Washingtonn D.C. to name a few populous 
and politically important places, whereas to the reader this is implied. 
  If this small a region cannot be mentioned it should be left out.

2) Eastern South America is not even defined in your figure.  Southern 
South America is, with a clear trend in Fig 3.14 (which is much weaker 
in the GPCC data), but this is not included in the SPM.

3) There is a clear trend in northern Europe, but as we all know that it 
is only in winter and the summer has in fact an opposite trend, would it 
be possible to add the word "winter"?

4) The North Asian trend.  Looking at the data from individual GHCN 
stations, almost all of them have lots of missing data around 1940, when 
the averaged series shows a big jump.  What is the evidence that this is 
not caused by chances in the observing system?  I find step function 
always quite suspect.  The VasclimO dataset, which the authors claim has 
better homogenization, has a decreasing trend for the period 1951-2000! 
  This does not seem the kind of certainty that warrants inclusion in 
the SPM.

5) I still see no significant trends in Central Asia except for 3 
stations in the far west of China and in Russia (see plot).  Do you want 
to make a sweeping statement "Central Asia is getting wetter" based on 
these three station series?  Wulomoi shows 1.5 decadal cycle that 
imitates a trend, Dulan and Irtyssk have barely significant trends 
(p=0.04).  There are many other stations with no trends.

6) I do not see an area labelled Sahel on your Figure in Chapter 3.  Why 
is it then included here?

The trend in the Sahel is only significant when you start late and 
finish early; rainfall has increased substantially again since 1995. 
Given the large decadal variability in the first half of the century, 
and the attribution to aerosols of the drought in the 1970s and 1980s, I 
would hesitate to call the remaining trend "significant".  Also, it is 
only the western Sahel that has a trend, not the eastern Sahel.

7) In the Mediteranean there is only a significant trend in North 
Africa, there is no significant trend on the northern shores.  The trend 
in the time series of Fig 3.14 is not very convincing by eye, it is much 
better if you take only the southern half, i.e., North Africa.  Claiming 
the "Mediterranean" is receiving less rainfall as a whole is again 
misleading.

8) From your plot (and mine on www.knmi.nl/adrica_cenarios) I see very 
strong decadal variability in southern Africa, and no significant trend. 
  We could just happen to have had a downward cycle near the end.  What 
value for the autocorrelation was used to determine the significance of 
the trend?  The judgement by eye agrees with the map, which does not 
show strong brown colours either.

9) From your map, this concerns Butan/Assam only; the rest of the 
subcontinent is getting wetter.  I see why the restriction on naming 
countries causes problems here...  In the GHCN dataset I find only one 
station with >70 years of data there with a significant downward trend, 
Darjeeling, and only a half dozen with >50 years between many more 
stations with no trend.  Again, you are basing a very important 
statement on very little actual data, and this statement will doubtless 
be interpreted to mean that large parts of teh subcontinent are drying. 
  I think it would be better to leave it out.

To my great surprise "Western Africa" is included in Fig. 3.14, with a 
steep decline, but this is not mentioned in the SPM!

Western Australia also shows up very clearly in the colours, but is 
ignored.  Why?  Because Giorgi used northern and southern Australia?

So, in spite of the background information I still do not understand how 
this statement follows from the observations.

Greetings,
	Geert Jan

</x-flowed>

Attachment Converted: "c:\eudora\attach\time series per yr0.png"

Attachment Converted: "c:\eudora\attach\Re Precipitation trends statem.png"

Attachment Converted: "c:\eudora\attach\time series per yr01.png"

Attachment Converted: "c:\eudora\attach\time series per yr02.png"

Attachment Converted: "c:\eudora\attach\Re Precipitation trends statem1.png"
