tag:blogger.com,1999:blog-28837843.post1264912905810552134..comments2021-10-18T09:59:29.048-07:00Comments on Quark Soup by David Appell: The CRU HackDavid Appellhttp://www.blogger.com/profile/03318269033139447591noreply@blogger.comBlogger31125tag:blogger.com,1999:blog-28837843.post-56256338654779509402009-11-26T00:12:21.465-08:002009-11-26T00:12:21.465-08:00(unscientific comment) So if a trade wind blows fr...(unscientific comment) So if a trade wind blows from the southeast it doesn't do that according to Navier-Stokes? How fairly interesting, in spite of all geostrophic winds from Antarctic High, may it live for ever.jyyhhttp://erimaassa.blogspot.comnoreply@blogger.comtag:blogger.com,1999:blog-28837843.post-64270458326107022992009-11-23T14:58:00.249-08:002009-11-23T14:58:00.249-08:00Aeronautical engineers work in thoroughly explored...Aeronautical engineers work in thoroughly explored regimes.<br /><br />That's where "pushing the envelope" comes from, actually testing something where you don't have data to see what happens.<br /><br />They know when they're doing it. It's an actual science.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-72069573620283664742009-11-23T14:53:01.180-08:002009-11-23T14:53:01.180-08:00I'm saying you can't solve the Navier Stok...I'm saying you can't solve the Navier Stokes equations, and you need to solve the Navier Stokes equations. So you can't predict the climate.<br /><br />A Kalman filter just maximizes the normal probability of the data given the parameters, as a way to determine the parameters. It happens to be organized to easily add new data.<br /><br />It's a way to formalize the model fitting as a statistical problem. It doesn't depend on whether there's physics there or not, just that there are parameters. You're not doing physics, you're doing curve fitting.<br /><br />Hamming: the purpose of computation is understanding, not numbers. Sort of the opposite of climate modelling.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-9499047480004646922009-11-23T12:48:08.628-08:002009-11-23T12:48:08.628-08:00> One of the advertised
> features of Kalma...> One of the advertised <br />> features of Kalman filters is the <br />> ability to determine parameters <br />> from data, no matter what the <br />> model is.<br /><br />Throwing around a big word like "Kalman filters" just convinces me more about that you're full of bullshit. I have yet to see any evidence that you know what climate modelers are actually doing, or that you've ever taken the time to read a document on the underlying design of a climate model. At best you've taken a course in PDEs and maybe introductory EE, about 30 yrs ago I suspect. You make a lot of big bullshit claims like you just need to solve the NS equations to determine future climate, which is clear utterly simplistic and something I'm highly sure physicists tried in about, maybe 1935. Stop giving this impression like you're somehow all-wise and the thousands of PHd climate scientists out there who work on this stuff their whole lives, 10 hrs a day are buffoons who can't even apply what they learned in sophomore calculus. <br /><br />If your methods are so much superior, where are your results (published or not)?David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-1102829664952158392009-11-23T12:40:47.362-08:002009-11-23T12:40:47.362-08:00> It's available to make the model
> th...> It's available to make the model <br />> that it's part of fit the data.<br /><br />Bullshit.<br /><br />Climate modelers don't choose whatever value of the effective viscosity is needed to somehow predict future warming. They use the values that scientists have measured for it, as a function of temperature, pressure, density, and/or other relevant parameters. This is true for all areas of applied physics, including ones you trust your life to. Do you think that aeronautical engineers, when they model the airplane wings we all ride all the time, calculate Reynolds numbers from first principles?David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-56498252511234713622009-11-23T12:38:01.660-08:002009-11-23T12:38:01.660-08:00> A parameter might be, oh,
> say "eff...> A parameter might be, oh, <br />> say "effective viscosity." <br />> V sub eff.<br /><br />So you complaint is that climate scientists don't determine effective viscosity from first principles, but instead use measured values?<br /><br />That's very, very lame. That scenario is true for every single branch of physics I can imagine. There are many, many fundamental parameters that physics can't calculate from first principles but instead use measured values.<br /><br />So what is your point?David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-68189501864326620822009-11-23T12:35:53.528-08:002009-11-23T12:35:53.528-08:00> Being a PDE doesn't make it
> underly...> Being a PDE doesn't make it <br />> underlying physics!<br /><br />Of course not. But almost all the relevant physics equations needed to model climate dynamics are PDEs -- the equations of fluid dynamics and heat conduction and thermodynamics and radiation transfers.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-40132627666884452752009-11-23T12:26:22.109-08:002009-11-23T12:26:22.109-08:00One of the advertised features of Kalman filters i...One of the advertised features of Kalman filters is the ability to determine parameters from data, no matter what the model is.<br /><br />Not that there's magic in Kalman filters, but it's a handy formal way to compute what could be done other ways.<br /><br />Feed in data and get out the best current estimate of V sub eff and all the other parameters; they can even be described by PDEs.<br /><br />That just makes explicit the direction the implications really go, from data to parameters, in models.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-40578059694725324742009-11-23T12:16:28.038-08:002009-11-23T12:16:28.038-08:00Being a PDE doesn't make it underlying physics...Being a PDE doesn't make it underlying physics!<br /><br />I pointed to the Navier Stokes equations as an example of what actually governs the problem, not as an example of a PDE.<br /><br />A parameter might be, oh, say "effective viscosity." V sub eff.<br /><br />Looks scientific, doesn't it.<br /><br />It's available to make the model that it's part of fit the data.<br /><br />This is not anything new.<br /><br />One interesting exercise is counting the degrees of freedom in stuff like that, and comparing how much data is being matched.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-56750339056523872342009-11-23T11:44:38.303-08:002009-11-23T11:44:38.303-08:00> I assure you they don't solve
> the N...> I assure you they don't solve <br />> the Navier Stokes equations<br /><br />That's not what I said.<br /><br />I said that they solve the underlying PDEs in exactly the same way you propose the NS equation be solved -- numerically.<br /><br />You talked like you had some new approach to the problem. But it turns out that all you meant was...to do what climate models are already doing.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-87459644398879498562009-11-23T11:40:58.689-08:002009-11-23T11:40:58.689-08:00> You model has parameters.
> You fuss with ...> You model has parameters.<br />> You fuss with the parameters.<br />> Then the model fits.<br /><br />Please tell us exactly what parameters modelers are "fussing with" to make sure their model fits the data.<br /><br />You can pick whichever of today's models you want.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-3461965410988418562009-11-23T11:25:34.169-08:002009-11-23T11:25:34.169-08:00There's lots of interesting natural physics.
...There's lots of interesting natural physics.<br /><br />I always liked the guess that the way wind produced large scale ocean waves was by maser action, with the smaller waves breaking perferentially on the tops of larger waves and thus adding energy in the right places.<br /><br />Whoever did that didn't have in mind funding for a bogus model. He was just curious where the hell the long waves come from, and thought this might be it.<br /><br />That's the kind of area where actual physics can do something.<br /><br />Model builders are a branching out towards a career path in management, as we used to call it.<br /><br />Funding is everything there.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-11806858884724022592009-11-23T11:09:14.721-08:002009-11-23T11:09:14.721-08:00You try your model out on past data, which it is n...You try your model out on past data, which it is no good if it doesn't fit.<br /><br />You model has parameters.<br /><br />You fuss with the parameters.<br /><br />Then the model fits.<br /><br />That's a statistical procedure; you can formalize it easily and do it automatically.<br /><br />It will work as well as a polynomial fit.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-78713594925973034762009-11-23T11:06:07.420-08:002009-11-23T11:06:07.420-08:00I assure you they don't solve the Navier Stoke...I assure you they don't solve the Navier Stokes equations.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-88878908457843034472009-11-23T10:38:03.433-08:002009-11-23T10:38:03.433-08:00> So you have to leave physics
> pretty muc...> So you have to leave physics <br />> pretty much behind. Enter model <br />> builders, unaffected by the these <br />> considerations. <br /><br />That's just absurd.<br /><br />Read a book like Pierrehumbert's "Principles of Planetary Climate," and you'll see some of the serious physics that form the basis for climate models. Read any of the documents describing a specific climate model, and you'll find it full of all kinds of physical equations.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-2618473936442777502009-11-23T10:25:14.784-08:002009-11-23T10:25:14.784-08:00> You can get nice numerical
> solutions of...> You can get nice numerical <br />> solutions of the Navier Stokes <br />> equations<br /><br />And what do you think a climate model is? <br /><br />It's numerical solutions to the underlying PDEs, including the NS equations. Except they incorporate a huge amount of other physics and physical processes that you can't get by just naively solving the NS equations.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-60528870163971091892009-11-23T10:22:04.431-08:002009-11-23T10:22:04.431-08:00> You can fit any past data easily.
Models do ...> You can fit any past data easily.<br /><br />Models do not "fit" data -- they predict values. My understanding is that most models today do not use so-called flux corrections (ref: IPCC 4AR WG1 Ch 8.2.7 p 607)--they are based purely on physics.<br /><br />> Models are an obscure version <br />> of a polynomial fit. <br /><br />That's wrong. They have their basis in physical laws, not statistics. The equations used are physical equations, not statistical equations. You should read one of the manuals that describes a model --David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-65295732311278835332009-11-23T09:15:34.870-08:002009-11-23T09:15:34.870-08:00"Sure, use the Navier Stokes equations."...<i>"Sure, use the Navier Stokes equations."<br /><br />Excuse me but what are those? Can they be solved in reasonable time when applied to a large and complex system?</i><br /><br />The core equations of fluid dynamics.<br /><br />They're hard to solve for several reasons, but they're the least that you have to solve.<br /><br />One reason being that disturbances in three dimensions cascade to smaller scales, so your grid will never be adequate for long.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-27399862935950453322009-11-23T08:32:47.712-08:002009-11-23T08:32:47.712-08:00> Threatening to delete data to
> avoid fre...> Threatening to delete data to <br />> avoid freedom of information<br /><br />Yes, that (if its true, and not added in by the hackers) is not acceptable.<br /><br />Most journals have (or now have) policies that require authors to archive their data someone and make them publically available. AGU journals do. Science does.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-57405697508527027092009-11-23T04:39:47.857-08:002009-11-23T04:39:47.857-08:00"Sure, use the Navier Stokes equations."..."Sure, use the Navier Stokes equations."<br /><br />Excuse me but what are those? Can they be solved in reasonable time when applied to a large and complex system? If not, how do you know they work in such a system? And please don't say they work because they work in smaller systems, since you cannot be sure. Why are you wasting everybody's time writing of something that you do not know if it works in such a system? Excuse me for the bluntness, but what are the error limits if you try to solve those in such a system?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-28837843.post-58525899191538400922009-11-23T03:35:37.143-08:002009-11-23T03:35:37.143-08:00Are you kidding? I've read them! Threatening t...Are you kidding? I've read them! Threatening to delete data to avoid freedom of information. We deserve a fully open public review of the raw unedited data to be looked over by the scientific community. If they have nothing to hide than this shouldn't be a problem. I do not deny "climate change" I do however know a thing or two about science and "tricking" data is a big no, no?. Denying access to third parties, shame on them. If I took massaged data to a client I could be charged for fraud. Editing data and deleting information. This is not unbiased science.Unknownhttps://www.blogger.com/profile/08033515859721973357noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-73685699520137098952009-11-22T18:33:42.410-08:002009-11-22T18:33:42.410-08:00You can fit any past data easily.
It has zero pre...You can fit any past data easily.<br /><br />It has zero predictive power, however.<br /><br />Imagine just running a polynomial through all the points, if it helps.<br /><br />It predicts the past perfectly.<br /><br />It predicts the future not at all.<br /><br />Models are an obscure version of a polynomial fit.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-60227108754096369762009-11-22T18:31:22.646-08:002009-11-22T18:31:22.646-08:00If you can't do it, you can't do it.
Ther...If you can't do it, you can't do it.<br /><br />There's lots of things you can't do.<br /><br />This is one.<br /><br />You can get nice numerical solutions of the Navier Stokes equations. It's well worth doing once, so you can get a feel for how far out of doable the climate is.<br /><br />"How far out" means huge orders of magnitude.<br /><br />So you have to leave physics pretty much behind. Enter model builders, unaffected by the these considerations.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-57962434322412583592009-11-22T17:32:01.080-08:002009-11-22T17:32:01.080-08:00People certainly started with elementary physical ...People certainly started with elementary physical models -- they predict warming.<br /><br />Solve the N-S equation? Sure. Go ahead and show us. They can barely be solved in simple fluid situations. In fact, it's not even known if they HAVE closed solutions in all situations in 3 dimensions.<br /><br />How are you going to factor in land use changes, other GHGs, ocean currents, evaporation, cloud physics, solar changes, aerosols, etc into the N-S equations? Even in principle?<br /><br />> Too hard to calculate with <br />> them? Pieces of information <br />> missing?<br />><br />> Right. You can't do it.<br /><br />Actually they can do it, as post-predictions of the 20th century show.<br /><br />In any case, just throwing up one's hands and saying "you can't do it" is not sufficient, as too much depends on the answer.David Appellhttps://www.blogger.com/profile/03318269033139447591noreply@blogger.comtag:blogger.com,1999:blog-28837843.post-67100159191827626242009-11-22T16:25:14.526-08:002009-11-22T16:25:14.526-08:00Sure, use the Navier Stokes equations.
Too hard t...Sure, use the Navier Stokes equations.<br /><br />Too hard to calculate with them? Pieces of information missing?<br /><br />Right. You can't do it.<br /><br />My own approach has always been Baysean.<br /><br />What are the odds that computers become widely available at exactly the moment a dire warning is needed?<br /><br />As opposed to, on the other hand, the Earth going along and making out somehow, under some sort of apparent stability, and thus not needing warnings.<br /><br />I'd suggest going back to more elementary physics, like what causes wind waves to grow, a hot topic in the 70s in the JGR as I recall, back when science was science.rhhardinhttps://www.blogger.com/profile/06901742898653890646noreply@blogger.com