tag:blogger.com,1999:blog-7729093380675162051.post5658589490959614454..comments2024-03-28T13:56:47.604+11:00Comments on moyhu: Another round of Pat Frank's "propagation of uncertainties.Nick Stokeshttp://www.blogger.com/profile/06377413236983002873noreply@blogger.comBlogger41125tag:blogger.com,1999:blog-7729093380675162051.post-31279735086664255092019-10-03T01:32:18.052+10:002019-10-03T01:32:18.052+10:00Hi,
The economics blog, EconLog, posted Pat Frank...Hi,<br /><br />The economics blog, EconLog, posted Pat Frank's theories with approval. I and others tried to point out some of the flaws. <br /><br />https://www.econlib.org/archives/2016/12/hooper_and_hend_2.html<br /><br />https://www.econlib.org/archives/2017/04/henderson_and_h.html<br /><br />One thing in particular that I pointed out was that satellite and balloon measurements also show warming. He was not persuaded (needless to say ;-)).<br /><br />Mark Bahnerhttps://www.blogger.com/profile/05768989010234358695noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-35095116234529842952019-09-25T05:06:07.942+10:002019-09-25T05:06:07.942+10:00"has "of a person" as the implicit ...<i>"has "of a person" as the implicit attached unit"</i><br />Careful, this is getting into Frank territory. And its not an SI unit.<br /><br />I think the right way to think of an average Av(x) is to perform a linear operation A(x), then divide by A(1), so Av(1)=1. Generally A(x)=∫w x dx, say, so you divide by ∫w dx. The units are those of x, noting extra implied. The extra unit introduced by integration cancels out. It's the same with summation.<br /><br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-39325217520276788192019-09-25T03:01:38.061+10:002019-09-25T03:01:38.061+10:00Another part of the difference is that baskets and...Another part of the difference is that baskets and apples are discrete quantities. You can have 1 basket; you can't have really have 2.3 baskets or 3.14 baskets. But temperature and time are continuous quantities. <br /><br />When you have an average over a discrete quantity (say, the average height of a group of people), it is implicitly about the unitted quantity. It is the height <i>of</i> a person. <br /><br />When you have an average of temperature over days, weeks, months, you're measuring temperature at given points in time, interpolating between those, and your average is the integral of this temperature function, divided by the time. Same rough idea, but discrete vs continuous. <br /><br />The same way that average height of a group of people has "of a person" as the implicit attached unit, an average temperature has "at a moment in time" as the implicit attached unit. Not over a month, or a year, but at an instant. <br /><br />Discrete vs. continuous.Windchasershttps://www.blogger.com/profile/11554275410734284781noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-14475287797004322652019-09-24T14:50:03.501+10:002019-09-24T14:50:03.501+10:00Piper,
I have had similar discussions, eg here, wh...Piper,<br />I have had similar discussions, eg <a href="https://wattsupwiththat.com/2019/09/07/propagation-of-error-and-the-reliability-of-global-air-temperature-projections-mark-ii/#comment-2790336" rel="nofollow">here</a>, where he said, yes, the annual average temperature at London would be properly given the units °C/year. Leter he tries to walk back.<br /><br />There was an interesting addendum <a href="https://wattsupwiththat.com/2019/09/19/emulation-4-w-m-long-wave-cloud-forcing-error-and-meaning/#comment-2800033" rel="nofollow">here</a> where I noted that his units were actually quite inconsistent. Despite all the table thumping, he more frequently writes the relevant quantity as ±4 Wm⁻² than ±4 Wm⁻² year⁻¹, and lately the year⁻¹ has been fading rapidly. Importantly, when he actually uses it, in Eq 5.2, it is ±4 Wm⁻², and wouldn't make sense otherwise. His reply is a doozy.<br /><br />With this stuff, you try to peel away one level of nuttiness, and there are whole new layers beneath.<br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-91943222613744768612019-09-24T04:29:16.484+10:002019-09-24T04:29:16.484+10:00In case anyone is interested, I've been discus...In case anyone is interested, I've been discussing this with Pat Frank over at PubPeer.<br /><br />**********************<br />Me:<br />**********************<br /><i>Let's say that I measured the temperature outside my house over the course of a year and the annual average was 10 C. According to you, that is actually 10 C/year.<br /><br />Now let's do a simple unit conversion. 10 C/year x 10 year/decade = 100 C/decade. According to you, the decadal average temperature outside my house is 100 C.<br /><br />That would be nuts.<br /><br />In the real world, averaging doesn't introduce any new units. You sum the measured quantities (preserving their units) and divide by the (unitless) number of measurements. Units remain unchanged.</i><br />**********************<br /><br />Pat's reply evades the "unit conversion" point I was making; instead, he turns to use of an (inappropriate) analogy:<br /><br />**********************<br />Pat Frank:<br />**********************<br /><i>In your example, your 10 C/year would merely imply that the next year average may be near 10 C as well.<br /><br />After 100 years, one could calculate a 1-year average from 100 years of data and see how closely your first year average came to the estimate from a larger data base.<br /><br />Let me make it easy for you with a more homely example. You own an apple orchard. After harvest, you have 100 baskets of apples. You count the apples and discover that the average is 100 apples per basket. Per basket, Piper. Averages make no sense without that 'per.'<br /><br />If we take your approach, we'd decide that after 100 years of harvests, every basket would include 100x100 = 10,000 applies. That's your logic. It produces utter nonsense.</i><br />**********************<br /><br />It seems to me that he really does not understand what he's saying. In Pat's example, if there is an average of 100 apples in one basket, then there would be an average of 1000 apples in 10 baskets. But that doesn't work with the temperature case -- if the temperature averages 10 C in one year, it does NOT average 100 C in 10 years.<br /><br />Thus my reply:<br /><br />**********************<br />Me:<br />**********************<br /><i>OK, let's apply the unit-conversion test to your apples example.<br /><br />You have measured an average of 100 apples in each basket. If we define a "decabasket" as 10 baskets, then the following can be used to extrapolate our estimate of the number of apples per basket:<br /><br />100 apples/basket x 10 baskets/decabasket = 1000 apples/decabasket<br /><br />That looks good to me. But now try it with the temperature example, using your claim that my front yard's annual mean temperature of 10 C is actually 10 C/year:<br /><br />10 C/year x 10 year/decade = 100 C/decade<br /><br />10 C/year x 0.083 year/month = 0.83 C/month<br /><br />So according to Pat Frank, the average temperature in my front yard is simultaneously 10 C ("per year") and 100 C ("per decade") and 0.83 C ("per month").<br /><br />Weird, huh? The answer is quite simple. Pat's "apple" example really IS about rates: the "average" that is being estimated is the average rate of apples per basket. But the "temperature" example IS NOT about rates. Pat doesn't understand this and so he turns it into a rate by appending a spurious "per year" unit to the quantity "10 C".<br /><br />The unit-conversion test is a handy way of demonstrating the problem with Pat's work here. If your work falls apart when you try to apply simple unit conversions, it means you're doing it wrong.</i><br />**********************Piper Chuarensenoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-86495286152644696362019-09-16T05:30:56.878+10:002019-09-16T05:30:56.878+10:00David Young, Hold my beer while I go look it up in...David Young, Hold my beer while I go look it up in my "LaPlace" transform tables, LOLpphttps://www.blogger.com/profile/15737287219806254245noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-11082323552340018672019-09-16T04:16:20.115+10:002019-09-16T04:16:20.115+10:00PP, You are of course denying the basic facts of ...PP, You are of course denying the basic facts of fluid dynamics here. All high Reynolds number flows are chaotic because of turbulence. There are also usually larger structures like vortex sheets/cores as well or convective shear chaotic shear layers. Slingo and Palmer are on very very solid scientific ground here. <br /><br />Your discovery of LaPlace's tidal equation is a nothing burger. Simple models sometimes work for chaotic systems provided they include key nonlinear elements. That's a head fake for the naive like yourself because these models only work in their calibrated regimes and sometimes not even there.<br /><br />You would learn more if you dropped the denial, the dissembling and the shameless self-promotion. If what you say is true, you could produce an ENSO forecest. You have never done so. Thus your assertions are just posturing so far as I can tell.David Younghttps://www.blogger.com/profile/17029429374522399227noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-16817660820774353492019-09-16T01:26:28.984+10:002019-09-16T01:26:28.984+10:00Slingo and Palmer are LOL hysterical. Reading tha...Slingo and Palmer are LOL hysterical. Reading that quote, it's clear that they're only saying their <b>simulation</b> is chaotic. This is the mistake you geniuses always make. A Lyapunov exponent can only be estimated for a mathematical formulation, but <b>never</b> on actual data. This means that it's very difficult to distinguish chaos from a complex yet non-chaotic waveform. <br /><br />To paraphrase Jesse Ventura, chaos is a crutch for the weak-minded pphttps://www.blogger.com/profile/15737287219806254245noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-70395261372936244522019-09-16T00:56:58.223+10:002019-09-16T00:56:58.223+10:00From Slingo and Palmer:
"Figure 12 shows 200...From Slingo and Palmer:<br /><br />"Figure 12 shows 2000 years of El Nino behaviour simulated by a state-of-the-art climate model forced with present day solar irradiance and greenhouse gas concentrations. The richness of the El Nino behaviour, decade by decade and century by century, testifies to the fundamentally chaotic nature of the system that we are attempting to predict. It challenges the way in which we evaluate models and emphasizes the importance of continuing to focus on observing and understanding processes and phenomena in the climate system. It is also a classic demonstration of the need for ensemble prediction systems on all time scales in order to sample the range of possible outcomes that even the real world could produce. Nothing is certain."<br /><br />This reflects what experts in fluid dynamics (in virtually all fields of application) say. David Younghttps://www.blogger.com/profile/17029429374522399227noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-39381818289701243282019-09-13T08:04:38.047+10:002019-09-13T08:04:38.047+10:00Wow, lots of ad homineming here. This site should...Wow, lots of ad homineming here. This site should be renamed "trashtalkers.com"<br /><br />See, I'd fit right in, but I gotta go now.Anonymoushttps://www.blogger.com/profile/00527645127036121231noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-49915063210377640772019-09-11T11:21:20.518+10:002019-09-11T11:21:20.518+10:00It would be very poetic - Et tu, Quoque?
Excellent...It would be very poetic - Et tu, Quoque?<br />Excellent critique of the paper, btw.Scott Simmonshttps://www.blogger.com/profile/07349460516783179575noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-1793619294739672452019-09-11T11:04:20.662+10:002019-09-11T11:04:20.662+10:00To quoque - if I ever feel the need of an alias, I...To quoque - if I ever feel the need of an alias, I'll use that.Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-73528191808982877032019-09-11T10:09:07.592+10:002019-09-11T10:09:07.592+10:00Sounds a little tu quoque-ish to me.Sounds a little tu quoque-ish to me.Scott Simmonshttps://www.blogger.com/profile/07349460516783179575noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-9936947804067375072019-09-11T02:14:35.237+10:002019-09-11T02:14:35.237+10:00If you're trying to measure something that'...If you're trying to measure something that's in W/m^2, and your uncertainty is in W/m^2/year, that means your measurements are getting worse each year. <br /><br />I.e., you say a forcing is 20 W/m^2, +/- 4. Your measurement is 95% likely within 16-24 W/m^2. <br /><br />If you add "/year", then that means next year, your estimate of forcing will be 20W/m^2 +/- ~7. And the year after that, 20, +-10, roughly. And so on. <br /><br />Does the uncertainty in the measurements grow each year? No, it definitely does not. The units here are incorrect. Windchasershttps://www.blogger.com/profile/11554275410734284781noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-23343803091795575802019-09-10T23:51:01.168+10:002019-09-10T23:51:01.168+10:00I try to transform the concept of error propagatio...I try to transform the concept of error propagation into the real climate system. So the development of cloud cover percentage would behave like a particle's dislocation due to brownian motion. Is that realistic? Of course not! Cloud cover is part of the global water cycle that takes about a week. When a cloud has rained out at a specific location, the "next" condensation process is not (or maybe only weakly) correlated to the state before. Also the aerosols have been washed out. There is no longtime mechanism of cloud cover accumulation. And the proposed idea of error propagation even allows negative cloud cover - that's totally unphysical. Cloud cover on a longer timescale is related to balance and not to random shifts.City Runnernoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-66741104423543940002019-09-10T15:08:32.833+10:002019-09-10T15:08:32.833+10:00"I think he could use 4 w/sqm monthly also, f...<i>"I think he could use 4 w/sqm monthly also, for 100 years divide by sqrt(1200) instead of sqrt of (100)."</i><br />Well, he could but he'd get a different answer. What is the use of an analysis that gives a different answer with whatever time period you choose, and your choice is arbitrary?<br /><br />Of course, the real answer is that he shouldn't be accumulating at all. There is no basis for doing so, and he doesn't give one. I'm just pointing out the contradictions in the approach, which are an adequate proof that you can't do it that way.<br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-21315859833069343742019-09-10T10:20:38.152+10:002019-09-10T10:20:38.152+10:00Sounds ok for that example. But power output is no...Sounds ok for that example. But power output is not an uncertainty. This is the main weak point in his analysis (which I still believe is correct). I think he could use 4 w/sqm monthly also, for 100 years divide by sqrt(1200) instead of sqrt of (100). The average uncertainty at any given time over all the model runs is 4 W/sqm. That is not much relative to solar irradiance (say 1360 W/sqm). You would want to use a reasonable timeframe, and 1 year is reasonable for 100 years of prediction. Since we want to propagate the uncertainty, we have to include the effect of i-1 in i.John Q Publichttp://johnqpublic.usnoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-61770971295323570392019-09-10T09:07:52.274+10:002019-09-10T09:07:52.274+10:00He's probably binning annually. That just mean...He's probably binning annually. That just means you collect annual subsets as your way of estimating something ongoing. It doesn't change the thing you are estimating, and it certainly doesn't change the units.<br /><br />Suppose you want to know how well your wind turbine is doing. So you calculate average output over week. 150 kW. Not 150 kW/week. Maybe try for a month. 120kW. Not 120 kW/month. Maybe put together 12 monthly averages. That's 130kW. Not 130 kW/month, because you binned in months. Nor 130 kW/year. That average is 130 kW/<br /><br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-76630480278455244842019-09-10T05:27:51.094+10:002019-09-10T05:27:51.094+10:00Re: Annual means (+/- 4 W/sqm)
I see that on page...Re: Annual means (+/- 4 W/sqm)<br /><br />I see that on page 3833, Section 3, Lauer starts to talk about the annual means. He says:<br /><br />“Just as for CA, the performance in reproducing the<br />observed multiyear **annual** mean LWP did not improve<br />considerably in CMIP5 compared with CMIP3.”<br /><br />He then talks a bit more about LWP, then starts specifying the means for LWP and other means, but appears to drop the formalism of stating “annual” means.<br /><br />For instance, immediately following the first quote he says,<br />“The rmse ranges between 20 and 129 g m^-2 in CMIP3<br />(multimodel mean = 22 g m^-2) and between 23 and<br />95 g m^-2 in CMIP5 (multimodel mean = 24 g m^-2).<br />For SCF and LCF, the spread among the models is much<br />smaller compared with CA and LWP. The agreement of<br />modeled SCF and LCF with observations is also better<br />than that of CA and LWP. The linear correlations for<br />SCF range between 0.83 and 0.94 (multimodel mean =<br />0.95) in CMIP3 and between 0.80 and 0.94 (multimodel<br />mean = 0.95) in CMIP5. The rmse of the multimodel<br />mean for SCF is 8 W m^-2 in both CMIP3 and CMIP5.”<br /><br />A bit further down he gets to LCF (the uncertainty Frank employed,<br />“For CMIP5, the correlation of the multimodel mean LCF is<br />0.93 (rmse = 4 W m^-2) and ranges between 0.70 and<br />0.92 (rmse = 4–11 W m^-2) for the individual models.”<br /><br />I interpret this as just dropping the formality of stating “annually” for each statistic because he stated it up front in the first quote.John Q Publichttp://johnqpublic.menoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-77901900488446567522019-09-10T05:04:13.086+10:002019-09-10T05:04:13.086+10:00IEHO a major impediment to rational discussion is ...IEHO a major impediment to rational discussion is the elevatin of a single parameter, global temperature anomaly, which although informative, only captures a small part of what we measure and know. <br /><br />Eli Rabetthttp://rabett.blogspot.comnoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-13374494562557583542019-09-10T01:26:11.529+10:002019-09-10T01:26:11.529+10:00The results are not chaotic, contrary to what this...The results are not chaotic, contrary to what this unpublished dude is saying. The variability in the climate signal is predominately due to ENSO and the other oceanic dipoles. These are equally as stable as the tides but only slightly more difficult to predict based on the tidal equations. <br /><br />pphttps://www.blogger.com/profile/15737287219806254245noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-90545455246638583042019-09-10T00:00:58.447+10:002019-09-10T00:00:58.447+10:00Well barry the results are chaotic. The climate p...Well barry the results are chaotic. The climate patterns are all over the place as Science of Doom's latest post demonstrates. The global average temperature anomaly is fairly consistent because of conservation of energy and tuning of top of atmosphere flux. Ocean heat uptake is usually roughly right too. But energy balance methods also get these things right. Everything else we really care about is all over the place.David Younghttps://www.blogger.com/profile/17029429374522399227noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-59977806358638406652019-09-09T17:11:27.139+10:002019-09-09T17:11:27.139+10:00Well I haven't followed every link, but I can ...Well I haven't followed every link, but I can only see a file of reviews of previous papers. I would be particularly interested in seeing what Carl Wunsch wrote, where can I find that pls?<br />Phil Clarkehttps://www.blogger.com/profile/15744659873337514317noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-12468780468137296932019-09-09T15:25:31.180+10:002019-09-09T15:25:31.180+10:00The reviews are linked in the WUWT article.The reviews are linked in the WUWT article.John_QPublichttp://johnqpubllic.usnoreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-85434861686809064382019-09-09T13:53:00.784+10:002019-09-09T13:53:00.784+10:00And I forgot to mention Frank's false forecast...And I forgot to mention Frank's false forecast of a continuing pause in global warming. Funny that he messed up on that, since he's now falsely claiming that GCM are largely useless for forecasts, even though they forecasted a continuing warming trend that he failed to:<br /><br />https://tamino.wordpress.com/2011/06/02/frankly-not/<br />https://web.archive.org/web/20180104193244/https://wattsupwiththat.com/2011/06/02/earth-itself-is-telling-us-there%E2%80%99s-nothing-to-worry-about-in-doubled-or-even-quadrupled-atmospheric-co2/Atomsk's Sanakanhttps://www.blogger.com/profile/06075386707195252260noreply@blogger.com