Pielke Comparison of Trends 1990-2007

Roger Pielke Jr reports that he has posted up a comparison of Hansen, IPCC, and observed trends 1990-2007 here:


and suggested that I start a thread (done!).


  1. Larry
    Posted Jan 18, 2008 at 1:10 PM | Permalink

    Link no work.

  2. Roger Pielke. Jr.
    Posted Jan 18, 2008 at 1:12 PM | Permalink


  3. conard
    Posted Jan 18, 2008 at 2:16 PM | Permalink

    a background / history of UAH data. link

  4. Mike B
    Posted Jan 18, 2008 at 2:59 PM | Permalink

    Dr. Pielke:

    You’ll find that the slope of scenario B is very sensitive to the starting point.

    The slope of scenario C is sensitive to both the beginning point and the ending point (especially if you go beyond2007).

    Scenario A’s slope is relatively insensitive to the starting and ending points.

    I see more value in comparing the predicted anomalies over the entire out of sample period (1988-2007). the anomales are far less sensitive to cherry-picking than the trends.

  5. Jim Arndt
    Posted Jan 18, 2008 at 3:00 PM | Permalink


    Why is the GISS so different from the others. Is the software used to filter the data different?

  6. Jim Arndt
    Posted Jan 18, 2008 at 3:08 PM | Permalink

    Hi Roger,

    Does your calculations filter out large events such as El Nino and Volcanism? What I am saying is that if you use them can they skew the result bias positive or negative?

  7. Roger Pielke. Jr.
    Posted Jan 18, 2008 at 3:21 PM | Permalink

    Mike B.- Slope for Scenario B starting from 1984 (when Hansen’s run began) is 0.24 (vs. 0.25 starting from 1990).

    Jim Arndt- I did no filtering, just took the projections as reported. I’d be interested in seeing why GISS is so different than the others. I’ve see claims that it is because of the extraoplations made in tghe polar regions, but given their size, seems implausible to account for such a large difference. It’d be nice to see GISS rsults with/without polar regions.

  8. Jim Arndt
    Posted Jan 18, 2008 at 3:30 PM | Permalink


    If I am correct isn’t the GISS a NASA data set and under Dr. Hansen’s direction. I may get a bad wrap for this but it does look like the data is skewed to the positive. I’m sorry but if this is NASA generated then they should have the raw data set some place for the GISS so it can be reviewed and filtered by other entities.

  9. Eric McFarland
    Posted Jan 18, 2008 at 4:04 PM | Permalink

    “The data for the Hansen scenarios was obtained at Climate Audit” You might want to consider posting the original source for obvious reasons.

    Also, Mark Bowen’s new book called “Censoring Science” gives some explanation as to how Hansen does his numbers. It’s a good and provacative read in all events.

    Meantime … FYI

    60 Minutes is re-airing its warming stories this Sunday … including the one about censorship and Hansen at NASA.

    Question: Why not have Steve do an in depth with 60 Minutes? Do tell …

  10. Eric McFarland
    Posted Jan 18, 2008 at 4:21 PM | Permalink

    Also, Hansen wants to know where his check for $720,000 from Soros is … he has yet to receive it:

    Click to access distro_Lawlessness_070927.pdf

  11. jae
    Posted Jan 18, 2008 at 4:22 PM | Permalink

    2. The outlier on surface observations, and the one consistent with Hansen’s Scenarios A and B is the NASA dataset overseen by Jim Hansen.

    Perhaps we just need to find Waldo.

  12. Roger Pielke. Jr.
    Posted Jan 18, 2008 at 4:22 PM | Permalink

    Eric- I added the direct link from the post. Thanks.

  13. Mike B
    Posted Jan 18, 2008 at 4:40 PM | Permalink

    Dr. Pielke:

    It would have been helpful had I been more precise:

    Using the RealClimate provided Hansen scenario data, here is the slope (deg C/decade) through 2007:

    start slope
    1984 0.248
    1985 0.236
    1986 0.232
    1987 0.234
    1988 0.236
    1989 0.248
    1990 0.266
    1991 0.293
    1992 0.318
    1993 0.359
    1994 0.405
    1995 0.444
    1996 0.488

    I would characterize these results as “slope is sensitive to start date”

  14. Mike B
    Posted Jan 18, 2008 at 4:41 PM | Permalink


    That was for Scenario B.

  15. Roger Pielke. Jr.
    Posted Jan 18, 2008 at 4:46 PM | Permalink

    Thanks Mike B. I certainly agree. I assume that part of the sensitivity has to do with the random volcanoes in B not found in A, hence the longer term views of B factor this out, hence the lesser degree of sensitivity starting prior to 1990. But I don’t think that this changes the conclusions that all Hansen scenarios lead to projected temperatures more consistent with IPCC 1990 than observations.

  16. Smokey
    Posted Jan 18, 2008 at 4:53 PM | Permalink

    It seems a little strange that James Hansen has specifically limited his disclaimer to one individual: George Soros. According to an Investor’s Business Daily editorial [The Soros Threat To Democracy, 9-24-2007], the money wasn’t paid by Soros directly; MovOn.org gave Hansen the money. If that’s not true, then Hansen may have a cause of action against the newspaper. But if it’s true, then Hansen’s best course of action is to set up a strawman – Soros, in this case, who funds MovOn.org – and argue that he hasn’t received any money from Soros.

    I would prefer to see Hansen state for the record that he has never received any money from any non-govenment groups, individuals or entities.

  17. Andrew
    Posted Jan 18, 2008 at 5:16 PM | Permalink

    Eric, by now no one is going to take Hansen at his word. They just aren’t. So if he says he didn’t receive money from Soros, or from the Heinz Foundation, I just don’t believe him. And if he claims to be censored….How can you possibly know? Wait….did he magically tell you with magical anti-censor pixie dust? If someone tells you they are being censored, call BS, becuase they shouldn’t even be able to say so. On topic, the IPCC appears to have based their projections on the more accurate satellite data. At least, those are closer trend wise. But here’s a question for Roger: can you give the trends at the time the predictions were made compared to what the predictions were? I have a theory I wish to test. That theory is that they based there trend for the future off of whatever trend happened to be going on at the time.

  18. Roger Pielke. Jr.
    Posted Jan 18, 2008 at 5:24 PM | Permalink

    Andrew- I discuss the notion of post-hoc curve fitting here:


    I suggested there:

    IPCC issued its first temperature prediction in 1990 (I actually use the prediction from the supplement to the 1990 report issued in 1992). Its 1995 report dramatically lowered this prediction. 2001 nudged this up a bit, and 2001 elevated the entire curve another small increment, keeping the slope the same. My hypothesis for what is going on here is that the various changes over time to the IPCC predictions reflect incrementally improved fits to observed temperature data, as more observations have come in since 1990…

    Imagine if your were asked to issue a prediction for the temperature trend over next week, and you are allowed to update that prediction every 2nd day. Regardless of where you think things will eventually end up, you’d be foolish not to include what you’ve observed in producing your mid-week updates. Was this behavior by the IPCC intentional or simply the inevitable result of using a prediction start-date years before the forecast was being issued? I have no idea. But the lesson for the IPCC should be quite clear: All predictions (and projections) that it issues should begin no earlier than the year that the prediction is being made.

  19. Bruce
    Posted Jan 18, 2008 at 5:25 PM | Permalink

    Soros funded a very controversial Lancet study on deaths in Iraq. The Lancet did not disclose it.


    It would be a surprise to me if Soros did not fund Hansen.

    As for GISS data – it is EXTREMELY suspect.

  20. Jim Arndt
    Posted Jan 18, 2008 at 5:28 PM | Permalink

    Hi Guys,

    Andrew I was thinking the same thing.
    Roger the IPCC seems to have adjusted their predictions to fit the time. It makes them seem more accurate. If you keep adjusting every 5 years or so then there is a good chance that your projections will fit the observed. See the following graph and the dates of the IPCC revisions.

  21. Sam Urbinto
    Posted Jan 18, 2008 at 5:43 PM | Permalink

    Everything I’ve read points to an organization that is super huge providing legal and media consulting to Hansen in regards to “censorship”. I would not think Soros directly had anything to do with it, and I didn’t see anything about money specifically. It’s a non issue anyway, who cares.

  22. John Lang
    Posted Jan 18, 2008 at 6:45 PM | Permalink

    Roger, I get 0.17C per decade with the RSS lower atmosphere data (versus 0.2C on your charts) and 0.18C per decade with the UAH data (versus 0.21C on your charts).

    Certainly makes some difference even with changes this small.

  23. John Lang
    Posted Jan 18, 2008 at 6:58 PM | Permalink

    I guess I should also add that the trendline per decade for RSS and UAH is rather meaningless.

    The January 1990 anomaly for RSS is -0.051C and the December 2007 anomaly is virtually the same at -0.046C. The UAH data shows an increase of 0.1C over the same period.

    The trend data is an artifact. Its like drawing a trendline over a half-circle.

    There is NO temperature change from January 1990 to December 2007. Your chart should really show 0.0C

  24. Posted Jan 18, 2008 at 7:26 PM | Permalink

    This shows anomalies with various parts of the Arctic (but not Antarctic) removed:

    Click to access ArcticEffect.pdf

  25. Andrew
    Posted Jan 19, 2008 at 9:54 AM | Permalink

    cce, the question then becomes, were the Arctic temperatures added properly? Maybe, maybe not. I’m sure this is debatable.

    Thanks, Roger, missed that bit.

  26. Posted Jan 19, 2008 at 11:55 AM | Permalink

    Two of the complications with the global temperature trend are ENSO activity and volcanoes. Here’s an attempt, with warts, to adjust the trend for ENSO and for the two major volcanoes. This uses 1979-2007: later I’ll look at 1990-2007.

    First, ENSO. I use satellite (RSS lower troposphere anomaly) data and the Oceanic Nino Index (ONI). I chose the period January 1984 – December 1990 (seven years) because it is between major volcanoes (El Chicon and Pinatubo) and because it is embedded in a longer period (1979- early 90s) of trendless (more or less) global SST.

    I checked temperature anomaly versus lag and got this plot . As expected, there is a lag between ENSO activity and its impact on global temperature. A five-month lag gives an r-squared of 0.43 (not great but its based on unsmoothed RSS data and it’s not bad for climate data).

    The plot for ENSO vs temperature anomaly is here . I used the slope coefficient to “remove” ENSO from the RSS data and got this plot . Note that the temperature change for this “de-ENSO’d” plot is 0.14C per decade.

    How does this compare with the unadjusted RSS data? Here is the unadjusted RSS data, which has a temperature change of 0.14 degree C per decade. It looks like changes in ENSO activity had no detectable impact on the 1979-2007 trend.

    Now, volcanoes. There were two major eruptions (El Chicon in 1982 and Pinatubo in 1991) in the satellite era. Note their clear impact on global temperatures for one to two years in this earlier plot .

    My wart-filled approach is to remove the temperature data following the eruptions (18 months for El Chicon and 24 months for Pinatubo) and replace it with the average of the 12 months immediately preceeding and immediately following the removed period. (I’m always open to ideas for better ways to remove their effects.)

    The result is here . The removal of the volcanic dips reduces the 1979-2007 trend to 0.10C per decade. I think that’s a fairly reasonable estimate.

    Note the pattern since the early 2000s, which suggests that “background” global temperature may have been flat to declining but has been somewhat “masked” by El Nino activity.

    As mentioned I’ll look at 1990-2007 and I may also estimate the one-time impact of the 1976 ENSO shift on global temperature.

  27. steven mosher
    Posted Jan 19, 2008 at 1:56 PM | Permalink

    re 26. remove the volcano effects by looking at the forcing profiles and back out the
    temp decrease.

    Crudely. Pinatubo had a -.2 W/sqm forcing. At .75C per watt, you can adjust the
    temp record and pretend that volcano never happened.

    Easy as pie.

  28. Peter D. Tillman
    Posted Jan 19, 2008 at 2:28 PM | Permalink

    #9, EDM, “Censoring Science”

    I can’t say I’m too impressed once I saw this blurb at Amazon: “Dr. Hansen and Science Meet President Bush and His Political Stooges!”

    Or this:
    “Now, for the first time and with unfiltered access, writer and physicist Mark Bowen finally tells the exclusive story of Hansen’s decades-long battle to bring the truth about global warming to light!” http://www.amazon.com/Censoring-Science-Inside-Political-Warming/dp/0525950141/

    Of course, this could be a balanced & unbiased treatment of Dr. “Death-train” Hansen…. ;-0

    Happy reading–
    Pete Tillman

  29. Roger Pielke. Jr.
    Posted Jan 19, 2008 at 2:52 PM | Permalink

    Over at RC Gavin Schmidt mistakenly claims that I have exaggerated the IPCC 1990 forecasts: “This is Pielke’s graph and the IPCC90/92 numbers are exaggerated by about 25% from what is actually in the 1992 report.”

    This is an complete misrepresentation, because he knows that in my analysis I rely on the 1990 report, not the 1992 report, partly in response to his complaints (and I know he knows this because we emailed about it):

    If Gavin Schmidt wants to know why some folks have lost respect for him, he can look no further than this sort of willful misrepresentation of another person’s work, which he has done several times over the past few weeks.

    Gavin then claims without offering any support that my “exaggerated” IPCC trends 1990-2007 “are still within the 2-sigma of the derived trends in the observations.”

    I responded by focusing on the 1990 report and the trends presented there. Lets see if they publish my comment (I’ll be surprised), and if Gavin offers any support for his claim (double surprised).


    Can you show your work to support your claim that the 1990 IPCC projection is “still within the 2-sigma of the derived trends in the observations”?

    The 1990 IPCC projection can be found here:

    1990 IPCC predicted trend = 0.33/decade
    Range of trend in 4 observational datasets (using GISS L/O) 1990-2007: 0.20-0.22

    In order for the IPCC trend to be within the 2-sigma of the observed requires that error to be > +/-0.13, which seems unreasonable given that the 2 sigma error in the 1979-2004 (only 7 years longer) is +/- 0.04 for the surface datasets and 0.08 for the satellites.

    What is the 2 sigma of the trends in observations 1990-2007?

  30. Posted Jan 19, 2008 at 4:12 PM | Permalink


    Gavin then claims without offering any support that my “exaggerated” IPCC trends 1990-2007 “are still within the 2-sigma of the derived trends in the observations.”

    I bet the trends are within 2 sigma for derived trends! Given the existence of large inter-annual autocorrelation (a part from the trend), the sizable variability of climate and the 0.05C known uncertainty in measurements for GISS land-based, they probably are within 2 sigma.

    That said: I haven’t done that calculation, and I haven’t seen it done. So, this is probably a place where it would be useful if someone did the calculation rather than arguing about it.

    In anycase, if the projections does fall inside the 2 sigma uncertainty for the slope, only saying we can’t falsify them. It’s not precisely saying the predictions are confirmed. (I haven’t figured out how to explain the problem I see with some of these statistical conversations. But it seems to me that, in terms of testing, the IPCC predictions have been made “the null hypothesis” as soon as they are created. That flips stastical tests over on their head.)

    It’s also valuable to show the projections did over predict the actual measured trends. IMO, when we look at these things, we need to acknowledge both things.

    Anyway, as I see it, both you and Gavin have some valid points– but there is a bit of a cherry pickers dream here.

  31. Roger Pielke. Jr.
    Posted Jan 19, 2008 at 4:19 PM | Permalink

    Lucia- Thanks. Gavin may indeed be correct, but it’d be nice to see him support his assertions with analysis rather than assertion, and of course there is that matter of misrepresenting my claims after I had updated my analysis , in part due to his complaints . . .

    As far as computing the significance levels of the temperature trends the CCSP report discusses this and presents trends for several time periods and measurements:

    Click to access sap1-1-final-appA.pdf


    Click to access sap1-1-final-chap3.pdf

  32. Roger Pielke. Jr.
    Posted Jan 19, 2008 at 4:51 PM | Permalink

    Gavin Schmidt responds, and tacitly acknowledges that he (a) misrepresented my work by ignoring my most recent presentation of IPCC 1990 predictions, and (b) that his claim that the2 sigma uncertainties in observations is smaller than the trend that I reported (note how he switches from my trend to his own, and ignores 3 of the 4 observational datasets). Even so, I’m pleased that he responded so that people can see for themselves.

    He also says some other things:

    Gavin Schmidt:

    [Response: To be clear, the graph that was linked to and that I was discussing was your original attempt at discussing the 1992 supplement forecast. The numbers on that graph showed 0.6 deg C for 1990 to 2007, which even you subsequently acknowledged (but never graphically showed) should have been 0.45 deg C. That implies a trend of 0.26 degC/dec. The trends in the two GISS indices over the same period are 0.22 +/- 0.09 degC/dec and 0.26 +/- 0.11 degC/dec (2-sigma confidence, with no adjustment for possible auto-correlation, using standard Numerical Recipes methods). You could play around with the uncertainties a little or pick different datasets, but the basic result is the same. Given the uncertainty in the trends because of interannual variability and the imprecision of the data, the 1992 forecast trends are within observational uncertainty.

    Now, for 1990 it is a slightly different matter. I would estimate the 1990 forecast is closer to 0.5 for 2007 (thus 0.29 degC/dec) than the 0.57 you graphed. It would then still be within the error bounds calculated above. For your estimate of the trend (and both of us are just reading off the graph here so third party replication might be useful at this point) it would fall outside for one index, but not the other. Making different assumptions about the structure of the uncertainties in that case could make a further difference in how it should be called. Since I wasn’t discussing this in the comment above, I’m happy whichever way it goes.

    For the record, your comments about this exchange in other forums are singularly unprofessional and rather disappointing. My only suggestions have been that a) you read from a graph correctly before making comparisons, and b) calculate error bars (as much as can be done) before pronouncing on significance. These might strike some as simple basic procedures. You appear to think they are optional. I’m not much worried that my professional reputation will suffer because of our apparent disagreement about that.

    One final note. I find the mode of discourse that you have engaged in – serial misquotation with multiple ad homs in various parts of the web with different messages for different audiences – unproductive and unpleasant. I am not the slightest bit interested in continuing it. – gavin]

    Here is my final response:

    Gavin- Thanks for reporting back. Thanks for observing that the trends that I reported for the 1990 IPCC are outside of the 2 sigma range. By picking your own smaller trend (rather than the one I reported) you came up with a slightly different answer. I see that you call for error bars when convenient (now) and ignore them when they are not (your comparison of Hansen’s 1988 projections with data). Of course, you well know that in my posts on this I had not claimed anything about statistical relationships of observed and predicted, simply presented the central tendencies.

    On your complaints, we can both feel misrepresented I suppose, and that is why it is best to present analyses when making claims, as especially when representing someone else’s work. So choose not to engage if you must, but conversations among people that disagree can can lead to constructive learning.

    Lets see if that appears. Rc seems to think that it is perfectly acceptable to make unfounded assertions about other people’s work, and let their commenters do the same, but gets all wound up when they are challenged in doing so.

  33. Posted Jan 19, 2008 at 5:44 PM | Permalink


    Thanks. Gavin may indeed be correct, but it’d be nice to see him support his assertions with analysis rather than assertion,

    Agreed. But, I cut everyone (you, Gavin, myself etc.) some slack about dotting i’s and crossing t’s in blog comments.

    and of course there is that matter of misrepresenting my claims after I had updated my analysis, in part due to his complaints

    Yes. But after discussion, he seems to have cleared that up. And I’d tack up some of the discord to the natural cross-talk that happens in blog comments.

    So, once again, bear in mind: Blog comments. There is a lot of cross talk in blog comments. And no matter how unpopular Gavin is here at CA, if you look through the thread, you’ll see he’s answering a heck of a lot of comments in a short span of time. The distinction of 1990 vs 1992 may be crystal clear in your mind, but it might not have been right there in the top of Gavin’s when he answered.

    I’m no apologist for Gavin (nor anyone else). I disagree with his assessment regarding the degree of wonderfulness of those Hansen 1988 predictions.

    But, I think we do need to cut people — including those with whom we sometimes disagree– some slack and not insist on the standards of peer reviewed articles in blog comments! (Yes, yes, in the interest of balance, I’ll close by saying, the RC guys should cut Steve M and you slack too. )

  34. Larry
    Posted Jan 19, 2008 at 6:05 PM | Permalink

    33, maybe we need peer-reviewed blog comments?

  35. Roger Pielke. Jr.
    Posted Jan 19, 2008 at 8:57 PM | Permalink


    I’d be cutting Gavin a lot more slack if (a) he and I hadn’t gone over this in detail in a lengthy email exchange last week, and (b) if this was the first time he or RC has misrepresented my work.

    Agreed not a big deal in the big picture, but it does become old after a while, especially his holier than thou repetition of how my efforts to correct his misstatements of my work is — in Gavin’s most recent words about me– “a poor reflection on your professional integrity”. When someone repeatedly mischaracterized another’s work and their “integrity” don’t be surprised to see a little return fire.


  36. Tilo Reber
    Posted Jan 20, 2008 at 10:54 PM | Permalink

    Maybe I’m missing something here, but are we giving Hansen 2 sigma uncertainty beyond the variation that already exists in the A, B, and C scenarios. If that is the case, it may be alright for scientific estimates, but it would seem to result in a range that is too broad for supporting political activism – which Hansen certainly engaged in.

  37. An Inquirer
    Posted Jan 22, 2008 at 3:14 PM | Permalink

    Certainly the GISS results raises questions (not only 2007 results, but also NYC’s Central Park adjustment and multitude of other issues), and when one calculates his own data to verify whether his model forecasts matches data, then red flags should be raised in the media, in Congress which funds the efforts, and even the head boss at NASA. My question: what auditing has been done or could be done of GISS methods to convert observed data to a single number? I have read Hansen’s documentation offered at RealClimate, and that is far from satisfactory. Has somebody gotten a hold of the algorithm used in the conversion process? Is this algorithm stable from year to year? And I suspect the algorithm cannot be completely identical year-to-year because the number of reporting stations vary.

%d bloggers like this: