xkcd – physicists

This cartoon, brought to my attention by John Havery Samuel, seems quite apt given what I sometimes try to do. Also, a little self-mockery is always a good thing.

credit : xkcd

credit : xkcd

Advertisements
This entry was posted in Comedy and tagged , . Bookmark the permalink.

31 Responses to xkcd – physicists

  1. Doug Bostrom says:

    The XKCD Mole of Moles model is worth a visit. Another spherical thought experiment. 🙂

  2. Doug, some of those are absolutely brilliant.

  3. Rachel says:

    That is one scary-looking mole.

  4. Rachel says:

    Very scary. Did they ever catch that one?

  5. Yup. The penultimate link in the “mole” tag list:

    http://climateaudit.org/2009/07/28/met-officecru-finds-the-mole/

    Twas the Auditor all along, sniffing the Interwebs.

  6. Rachel says:

    But I didn’t think he was the hacker. They didn’t actually catch the person responsible did they?

  7. BBD says:

    And here’s me thinking it was a criminal act – theft by an outside agent, aka “hacker”. Indeed, such is the conclusion of the police investigation. Odd that SM should spend so much energy on speculative misdirection. Nice to see Mosher (as ever) in comments, aiding and abetting.

  8. BBD says:

    Wotts – OT but let us know if Lomborg gets back to you about that borked graph he keeps waving around.

    Is “embarrassingly wrong” uncivil, btw?

    🙂

  9. He hasn’t every other time I’ve pointed it out, so I’m not hopeful this time. I didn’t regard “embarrassingly wrong” as uncivil. Seems like a reasonable description of what he’s done there, in my opinion at least.

    Are you lurking on Twitter, by the way?

  10. Rachel says:

    He probably doesn’t understand why it’s wrong.

  11. BBD says:

    @ Wotts – not lurking – I saw it on the Twitter sidebar here. I wasn’t entirely serious about the incivility btw – perhaps you can imagine how I might describe that graph? Though not on your blog, of course 😉

    @ Rachel

    I’d be very surprised if Lomborg didn’t realise how misleading that graph is. He’s been dealing in misrepresentations for so long it is virtually inconceivable that he could be unaware of what he is doing. And even if he did somehow make an innocent botch of it, lots of people, including Wotts, have since pointed out the problem out to him. His unresponsiveness damns him.

  12. BBD, yes I forgotten about that. Rachel, it is such a trivial mistake to make that if he hasn’t acknowledged it by now it’s really hard to imagine that he doesn’t realise. Either that, or his mathematical abilities are vanishingly small.

  13. Rachel says:

    Nobody ever went broke underestimating the intelligence of the American public.

    I’m sure it applies to all nationalities.

  14. Rachel says:

    Hmm. Do you think my comment is a bit uncivil? Perhaps it should be removed?

    I guess I would rather think that Bjorn Lomborg doesn’t understand something than that he is being deliberately deceptive.

  15. guthrie says:

    What is wrong with the graph? THe starting 3 year baseline seems a bit short, which no doubt makes some cherry picking, and it’s not got any error bars. That’s all I can spot at first glance.

  16. guthrie says:

    Never mind, I found your post on the topic. Look first and post later…

  17. Tom Curtis says:

    guthrie, the graph uses (effectively) a weighted 13 year baseline by base-lining an eleven year running mean over a four year period. Only the central four years have maximum weight. By including an 11 year weighted average, Lomborg manages to include the 1977, the 1987, and of course, the record breaking 1982-1983 El Nino events. The 82-83 El Nino still remains the strongest El Nino in 140 years of data using the SOI. The inclusion of three El Ninos with no La Ninas (except the tale end of the 74-75 La Nina with very low weight) biases the observational record upward during that period so that, by base-lining against that period, it shifts the model data upwards with respect to the observational data in all other periods. This bias may account for half or more of the model/data discrepancy in Lomborg’s figure.

    Making it worse, it is transparent that the satellite data (which starts in 1979) cannot have a 11 year mean with a central point in 1979, nor a central point in 2013. Lomborg appears to have used a 5.5 year mean for end years, and increased the interval progressively until an 11 year mean is achieved. This very strongly weights end points, and in particular the record breaking 1882-1983 El Nino at the start of the record, and the record breaking 2011-2012 La Nina at the end of the record, not to mention the strong 2008 La Nina. This is almost entirely responsible for the satellite record underlying the surface record for almost its entire length. The diminishing mean is used for observational records as well, and explains a further significant fraction of the difference between the models and observations.

    To make it worse still, Lomborg only shows an eleven year running mean which suppresses short term variation, thus concealing the fact that short term variation accounts for much of the difference; and further, suppresses the error margins (as you note) which also hides the fact that observations remain within the 95% confidence interval of the models (though only just).

    The observations are in fact running about 26% below AR4 models, even allowing for ENSO. Expressed differently, the AR5 models are running about 35% above observations. For AR4, the respective figures are 15% and 17.6%. This is partly accounted for by slightly increased volcanic activity not included in the models, and to a much lesser extent, by increased aerosol emissions by China and India. It is also partly accounted for by reduced solar activity.

  18. Tom Curtis says:

    I should add to the above, that:
    1) Satellite data is more strongly effected by ENSO events than are surface data, strengthening the effects discussed; and
    2) Lomborg chooses the RCP 2.6 scenario which has a very strong initial trend due to the rapid reduction of aerosol emissions, even though it does not warm as much as other scenarios in the long term. The initial trend of RCP 2.6 is greater than that of RCPs 4.5 and 6, and about the same as RCP 8.5. Thus, while givign the appearance of being conservative, Lomborg is actually comparing observations to a worst case scenario (for the short term).

  19. Rachel says:

    The 71% on Bjorn Lomborg’s graph is wrong. It depends entirely on where he starts his graph. If he chose a baseline further in the past – imagine shifting the y- and x-axis down and left – then he’d get a much smaller percentage. If he chose a more recent baseline – imagine shifting the axes in the opposite direction – then he’d get a huge percentage. It’s a meaningless number and if he knows this, he’s being deliberately deceptive.

  20. BBD says:

    As I said above – and in the light of Tom’s excellent recap of the many manipulation carried out here – how can Lomborg *not* be aware that this graph is confected and misleading?

    And then there is his flat unresponsiveness to detailed critiques of what he has done. This makes it all but certain that he is acting in bad faith.

  21. Tom, when you say

    The observations are in fact running about 26% below AR4 models, even allowing for ENSO. Expressed differently, the AR5 models are running about 35% above observations. For AR4, the respective figures are 15% and 17.6%.

    are you measuring this relative to the same baseline as Lomborg? As much as I agree with everything you’ve said about this figure, my very simple criticism is that it’s relative to his chosen baseline, so – it seems to me – that saying “models are 71-159% too hot” isn’t really a meaningful statement. I guess, one could be very explicit about the chosen baseline, but even then it seems like a rather odd way to compare the models and observations.

  22. Tom Curtis says:

    Wotts, the AR5 multimodel mean trend for the period 1990-2020 is about 0.23 C per decade. For AR4 the equivalent figure is 0.2 C per decade. In contrast, for Foster and Rahmstorf (2011), the adjusted trend from 1979-2010 is approximately 0.17 C per decade. There is a little acceleration in the Foster and Rahmstorf figures, as shown on the SkS trend calculator (Note, CRU = HadCRUT3). From 1990 to current the ground based trends are around 0.185 C per decade, and around 0.17 for satellite. That varies, generally downward, depending on the start year. Further, the Foster and Rahmstorf adjusted data is adjusted for volcanic and solar forcings, which are included in the AR4 models (up to 2000) and AR5 models (up to 2006). Eliminating those adjustments would drop the trend back towards the 0.17 C long term adjusted trend, which I use for the calculations.

    Further, when trends are calculated for ENSO equivalent years (excluding those effected by volcanoes), trends also come out around the 0.17 C per decade mark.

    Because I am comparing trends, baseline is irrelevant although start point is not. Unfortunately, no study shows the ENSO adjusted data alone, and I have not attempted to calculate it from the F&R values myself, so the 0.17 C per decade is imprecise – but close enough in my opinion to be confident that models are currently over predicting temperature increase. But by a small amount, and even that small amount may only be due to overpredicting changes in forcings.

  23. Tom Curtis says:

    Rachel, if he shifted his baseline to the right (more recent), that would both shift the model projection higher relative to the observed data (up until the crossover point around 1998), and decrease the absolute value of the observed increase, both of which would increase the percentage difference. Shifting it into the past would raise the relative position of the observations, and increase the absolute value of the observed increase, both of which would lower the percentage difference. You are correct that it is a meaningless number, but the equivalent comparison of trends is not.

  24. Tom,
    I did take FR11’s R-script into pieces and computed the individual contributons separately a while ago. Basically, I was using the same mutliple regression, but only adjusting the original data with the coefficient for MEI (as ENSO equivalent). Right now, I have the plots for GISS and HadCRUT4 available. However, they are not yet updated with 2012 data, which will bring the trend down a bit. I’m gonna update them as soon as I find the time: [GISS MEI adjusted] and [HadCRUT4 MEI adjusted]
    Hope that helps.

  25. Tom Curtis says:

    Thanks K.a.r.S.t.e.N. From your charts, you have an MEI adjusted GISS trend from 1979-2011 of 0.178 C per decade, and a MEI adjusted HadCRUT4 trend of 0.183 C per decade. Adjusting my estimate of the ENSO adjusted observed trend to 0.18 C per decade based on this would result in AR4 models showing trends 11% above observations, and AR5 showing trends 28% above observations.

  26. Tom, thanks, I hadn’t appreciated you were comparing trends. I should probably have realised 🙂 I agree, comparing trends is fine. Doing what Lomborg has done, though, is largely meaningless.

  27. KR says:

    Wotts – In light of the cartoon theme of this thread, and some recent issues with comments, the following Scene From A Multiverse may be of interest and relevance. Or at least a chuckle…

  28. Tom, here the update of FR11 for GISS, HadCRUT4, UAH. I have included 2013 with a simple interpolation for the missing 2 months of this year:
    GISS MEI adjustedHadCRUT4 MEI adjustedUAH TLT MEI adjusted
    Replace “MEI” with “ALL” in each link and you get to the all adjusted version (Solar/Volcanoes/MEI). Images will be updated as new data are available (hence subject to change).

    Interestingly, one Troy Masters is trying to discredit FR11 in a rather unconvincing fashion. His attempt has just been accepted for discussion at ESDD. Altough I am considering it an interesting effort (albeit fairly well covered in the literature already), I have my doubts that he got it right (for several reasons, which to explain goes far beyond a blog comment). Perhaps I’m gonna submit a comment if time allows (as it requires some efforts to disprove him). But anyways, you might have a look for yourself. Perhaps you are more impressed 😉

  29. Karsten, thanks. Very interesting.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s