The Blackboard

Just as a side note, the anomaly figures above are based on the 1200 km smoothing/data-infilling process. If 250km smoothing is used, the anomaly is still 39.

Steve McIntyre has a subject on changing rewriting history time and time again:

http://www.climateaudit.org/?p=2964

This is just more of the same

Fred–
They do have a variety of products, don’t they? This is the set I always use. It’s the one linked here:
http://data.giss.nasa.gov/gistemp/

And found under the heading “# 1880-present, updated through most recent month
# Global-mean monthly, annual and seasonal land-ocean temperature index, 1880-present, updated through most recent month”

I figure that since it’s the one GISS actually references at their temperature page, it’s what they consider “most standard”.

BobB–
These changes aren’t like the more dramatic episodes SteveM has discussed. It’s just a bit disconcerting to see temperatures as far back as 2005 changing, and knowing it’s likely not because they found a mistake in the 2005 data. It’s their method.

Ordinarily, the method of measuring something that happened in Nov 2005 doesn’t depend on what happened in 2008. But with GISSTemp, it does.

Anyway, this is also the reason I often state caveats like… my hypothesis tests are contingent on the data not changing!

“Most standard” or “most likely to give a more positive anomaly”? It is understandable that this is the one that GISS publicly shows. It gives the impression that the temperature dataset has great global coverage, has up-to-date information and shows the horror of AGW. But in reality, when one digs a little deeper, there are vast tracks of land (eg most of Africa) for which there is no data. Zip. Zilch. Nada.

Fred–
Yep! As much as people like to argue about whether GISS is better than HadCrut due to supposedly dealing with the poles, the issue is whether one likes more or less interpolation to deal with gaps in coverage. All the surface measurements have big gaps in coverages. The Troposphere is not the surface. So…which do we compare to “about 2 C/century” (which, trend I have noted before is rounded down from the values in tables in the AR(4) )

I’m sure it been figured out before but the co-relation (base period issues aside) between the satellite datasets and the surface datasets is fairly strong (right term?). But it isn’t perfect. I would hazard a guess that the variation differences between the two types are more due to the missing data and/or the infilling processes than surface vs troposphere differences… But I could be wrong.

Fred– The correlation between the surface and satellite sets is strong. Atmoz has posted about it. I’ve seen it in my numbers.

Of course it’s not perfect. But then the correlations between HadCrut, NOAA/NCDC and GISS GMST anomalies are also imperfect.

Whether the difference is more strongly affected by infillig or missing data I don’t know. If we are lucky, the IPCC will publish clear projections for the Lower Troposphere in the Fifth report.

But am I correct in assuming that the corelation between surface (and satellite) datasets is stronger than between surface and satellite ?

The correlation between “surface A” and “surface B” is stronger than the correlation between “satellite A” and “surface A”. This is what we’d expect both because surface should be similar to surface and because the surface sets themselves share some common station measurements etc.

So not only is there teleconnection, there’s “temporalconnection” in the GISS data!

I don’t get it but I was looking at the cache vs current and there are changes here and there throughout going all the way back to 1881 Aug, which changes from -13 to -12.

At the bottom our your post you have the following:
Reference: Current GISS Land/Ocean anomalies
Link goes here: http://hadobs.metoffice.com/ha.....sh/monthly

This is the HadCRUT3 surface temperature anomaly data if I am not mistaken.

Maybe I am just not smart enough to understand this, but does anyone else think that this adjusting of the historic temperature data all the way back to 1881 is just plain wrong? I guess righteous indignation is not PC these days.
Sometimes I feel like I’m in some kind of strange alternate universe.

JoeH– Thanks! Fixed it.
I know I was checking HadCrut too. I keep hoping they’ll release their data too.

Mike Bryant– I sort of understand the reason for these backward adjustments. But the fact that changing data for this year can propagate changes all the way back to 1881 makes it clear that “data” aren’t purely observations. The change from -13 to -12 is small though– those are in 100ths of a C.

Small or large, rewriting history is still wrong. Or perhaps I’m wrong and many, many small changes is just ducky. Talk about death by a thousand cuts, (and increases too.)

Broken record, Who, What, Where, Why and When. Version control. It has been said that professional software people do it that way. So many people wondering why something has changed. Maybe sloppy, maybe good, maybe ?
Do it by the book ,then less work more understanding. Data and Programs both. For money transactional with full audit and reconciliation.

Lucia: I believe GISS added missing Antarctic data in the interim. I don’t recall it being present on this map a month ago. Sometimes it’s there when they release the original monthly data, and sometimes it’s not. Did anyone save a copy of the old August map or can you retrieve an old copy?

http://data.giss.nasa.gov/cgi-.....mp;pol=reg

Considering the magnitude of the rise, the Antarctic would be the place to look.

Changing the GISS historical records to me is very suspicious. Is the base line also moving around?

To compound the problem, it would appear that some data fiddling also occurs at the ground station level. I have been watching one particular site for some years in Central West NSW. If the average temperature for the month is running minus then an odd warm day, of between 5 to 10 degrees above the actual, will some how sneak in.

Give me the Satellite data any day.

GISS/NASA employees do not report the data.

They instead report the results of application of a model to the measured values.

Additionally, that model includes a temporal teleconnection aspect in that future physical phenomena and processes will affect past physical phenomena and processes.

Some kind of an extension of quantum theory to climate, I guess.

Looks like our conversation caused Pielke Sr. to post: http://climatesci.org/2008/10/.....anomalies/

I didn’t save August’s GISTemp map, but if memory serves me correctly, there was little to no Canadian stations as well. A few of those have been added in the interim as well.

[...] we all know, the GISS Temp for August 2008 rose during Sept 2008. After the update, the GISS Temp for Sept 2008 was lower than the August 2008 temperature. Are you [...]

From Fred Nieuwenhuis (Comment#5714) link

Lucia,

This seems to imply that at least some of the anomaly data sets are not created by adjusting each months temperature relative to the same months anomaly baseline (e.g. Sept temp vs all other Sept of the baseline) but relative to all the months in the baseline period (all Jan-Dec temps of the basline period). Hadley seem to work with anomalies from each site so it wouldn’t correct for them. However I thought your assumption was that the anomalies were a month to month thing and not what Pielke is implying. Would this affect your analysis?

Thus the change in an anomaly value between months must be placed within the context of its percent change over this time period relative to the average expected change. For September, for example, a change of an anomaly from August to September of 0.1C is a 17% variance from the average expected change. A positive value (such as reported for this past September) means the global cooling was less than the average cooling rate for this month.

I think Roger Sr. is comparing anomalies to non-anomalies.

If the anomaly increases +0.1 C, the non-anomaly changed 0.1C less that it usually changes.

It’s like this (with made up numbers). Suppose the “true-non-anomaly” Northern Hemisphere Mean ST for August averaged over the 30 baseline years is 15C. Since fall is coming, the temperatures fall in the fall, and now suppose the “true-non-anomaly Mean ST for September averaged over the 30 baseline years is 14C.

Now, take year “N”. Suppose the average ST in August is 14.9C. That’s an anomaly of 14.9C-15C = -0.1C . Now, suppose the temperature in September drops to 14 C. The September anomaly is 0C. So, the anomalie rose 0.1C. But, that rise represents 10% of the real change we expect for September!

(Roger’s discussion uses real numbers, so he says a 17% variance.)

So, it’s not surprising these anomaly values are darn variable. They are based on underlying real temperatures which themselves display an annual cycle.

I think what Pielke is describing is the way I interpreted the anomalies — so no effect. The monthly anomalies are relative to the average for that month.

Lucia,
Thank you for the explanation… I think.
Mike

lucia (Comment#5695) October 7th, 2008 at 1:08 pm
> Anyway, this is also the reason I often state caveats like… my
> hypothesis tests are contingent on the data not changing!

Then forget GISS entirely. There’s a geek website http://www.theregister.co.uk which occasionally writes about climate stuff. If you think what you’ve seen so far is bad, take a look at http://www.theregister.co.uk/2.....rmometers/ Read it and weep. A couple of quotes follow. Note that the article seems to use ye olde Fahrenheit scale, rather than Celsius.

> Particularly troubling are the years from 1986-1998. In the
> 2007 version of the graph, the 1986 data was adjusted
> upwards by 0.4 degrees relative to the 1999 graph. In fact,
> every year except one from 1986-1998 was adjusted
> upwards, by an average of 0.2 degrees. If someone wanted
> to present a case for a lot of recent warming, adjusting data
> upwards would be an excellent way to do it.

> Looking at the NASA website, we can see that the person in
> charge of the temperature data is the eminent Dr. James
> Hansen – Al Gore’s science advisor and the world’s leading
> long-term advocate of global warming.

and the second quote…

> NASA has a very small number of long-term stations in
> the Arctic, and even fewer in Africa and South America.
> The data has been systematically adjusted upwards in
> recent years – as can be seen in this graph, reproduced
> below. Temperatures from the years 1990 to present
> have more than one-half degree Fahrenheit artificially
> added on to them – which may account for most of the
> upwards trend in the NASA temperature set.

The graph they refer to is a copy of a graph at http://cdiac.ornl.gov/epubs/nd.....raw_pg.gif

The monthly adjustments add up over time. Using archive.org (a.k.a. “The Wayback Machine”), I dug up a few copies of GISS all the way back to late 2005. I ran a quick-n-dirty comparison between that data, and the same period from the March 2008 version. In less than two-and-a-half years, the accumulated changes included stuff like…

- the average temperature from 1880 to the end of 1920 went down 0.03382 C.

- the average temperature from 1970 to the late 2005 went up 0.00981 C.

Guess what *THAT* does to the trendline? The only real Anthropogenic Global Warming seems to come from the keyboard of Dr. Hansen. And if Hansen accomplished that in 2.5 years, what has he done to the data in the last 25 years???

Walter Dnes: The referenced graph is for USHCN, which is for the United States, not the GHCN, which is global. The use of that USHCN graph by the author of your linked article during discussions of the Arctic, or South America, or Africa, therefore, is in error. Based on your quote, he is attributing that graph to NASA, when in fact it is the output of NOAA, particularly the NCDC. Refer to:
http://www.ncdc.noaa.gov/oa/cl.....ushcn.html

Lucia,
I’m surprised that you are surprised by this. I suppose it is important to keep pointing it out. My favorite example is how US temperatures around 1890 dropped sharply in Sept 2007, (shortly after Steve Mc found the bug around the year 2000). So the US warming trend increased significantly overnight. In August 2007, GISS US temp was
1885 -.50
1886 -.25
1887 -.21
1888 -.28
1889 .28
1890 .23
But in Sept 2007 it became
1885 -.65
1886 -.42
1887 -.25
1888 -.42
1889 .23
1890 .10

Paul– I’m not surprised. The monthly temperatures for GISS change all the time. But the revision for August is fairly large compared to the normal monthly revision.

FWIW, I think it’s utterly normal for any of the agencies to update month”N” when they report “N+1″. Errors are going to happen. On the other hand, it is weird to watch the temperatures for the 1880’s.

I’m aware of the error SteveM found. The NASA GISS bungle caught by SteveM is embarrassing for them. Hansen responded less than graciously.

Hadcrut is out: 0.376
http://hadobs.metoffice.com/ha.....sh/monthly

[...] according to GISSTemp, August’s anomaly dropped 0.11C during Setember. The change was sufficient to make it difficult to say whether the GISSTemp Sept. anomaly was hotter [...]