[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

*To*: dinosaur@usc.edu*Subject*: Numerical Models and Running Rexes*From*: MariusRomanus@aol.com*Date*: Sat, 21 Sep 2002 01:17:39 -0400*Reply-to*: MariusRomanus@aol.com*Sender*: owner-dinosaur@usc.edu

Right now I believe that I am in a particularly good position that allows me to make a few short comments about numerical models. In my atmospheric dynamics courses, Iâve been constantly learning about how numerical models for climate data modeling and weather prediction are made, how they function, and what they are able to reliably produce for describing the dynamical workings of the atmosphere. Why should this matter to the DML??? Simply put, Iâm learning just how ârealâ numerical models of natural entities truly are. There is a profoundly important concept that every single one of my professors continually mentions in regards to ANY type of numerical model... For even the best of the best, in order to "get it right", numerical models are composed of a core of respectability, surrounded by a varying amount of voodoo (fudge factors). General, broad-sweeping numerical models have a very high degree of confidence for their outputs. Numerical models designed to interpret the over all global climatic effects caused by glaciations, are just swell. The driving variables of these numerical models are all well known and can be independently verified using numerous different approaches, including looking at deep-sea cores to measuring the albedo from the Antarctic ice-sheet. On the other hand... Devising a numerical model that has a fairly high degree of accuracy and describes the "weather" caused by glaciation events, is not so swell. Iâm sure you understand why. (Remember, climate is the synthe! si! s or "statistics" of weather in a particular region. Weather is the "instantaneous condition" of the ocean and atmosphere at a "select location within" a region.) Look at it this wayâ. There is a reason why meteorologists can tell you that there are going to be thunderstorms across the entire state of Ohio todayâ. but they canât tell you that the storm presently at grid-point A9 is going to be a storm that is going to produce an F3 tornado once it moves into Mahoning County at 4:17pm. So where am I going with this? What I am getting at is simple. When you are dealing with numerical models thatâs purpose is to describe some specific, NOT general, natural entity, the subsequent loss in generality of the inputs results in the loss in accuracy of the outputs. I know this is confusing, but follow me here... When you start to become specific in numerical models, unknowns, already known and unknown, amplify already understood uncertainties, which in turn amplify already understood errors, while at the same time introducing brand new errors, which are also known and unknown. These now amplified errors are compounded beneath the fabric of your now "detailed" variables and uncertainties, both known and unknown. Error is built on top of error. The end result is a numerical model thatâs output confidence has been rapidly eroded away. How do you deal with all of this? Voodoo is the answer... Simplify your equations. Simplify your constraints. Simplify your variabl! es! . Approximate. Atmospheric dynamics is governed by only a handful of equations. Problem is, the equations are unsolvable. How do you get around that little bump in the road? Linearize them. Too bad that what you get out can model an air mass but canât model a storm within that air mass. And you know what else helps? Being able to look at your subject and test the accuracy of your numerical model against it. Itâs fantastic that Mr. Hutchinson was able to produce an accurate numerical model of an ostrich. Problem is, tyrannosaurs were not ostriches, just as a hurricane is not a extratropical cyclone, even though both are powerful storms with very similar structure. Tiny differences end up making a world of difference. (And with a bit of intentional imagery, one even evolves from the other.) So, it rightly follows that using the same numerical model to describe both of them will produce erroneous results. Not completely wrong results.. but results in error. Then why is it that a numerical model with set parameters, variable constraints and sensitivity analysis ideal for describing the locomotory capabilities of an ostrich, is thought of as being the numerical model that is ideal for describing the locomotory capabilities of a different, yet similar animal, n! am! ely a tyrannosaur? As Mr. Hutchinson said, "Our model included all of the tyrannosaur anatomy that was needed, and T.rex was still found wanting." Mr. Hutchinson also said that anatomists suffer from tunnel vision... They are mired in "untested assumptions about the relationship of anatomy and locomotor performance." Mr. Hutchinson says, "I think biomechanics is what is needed to test anatomical specializations and see how important they really are; the work has not been done to my satisfaction." And finally, Mr. Hutchinson says, "But computer models will always be a useful supplement to any other line of evidence, particularly when it comes to extinct animals, because they allow you to test hypotheses indirectly without relying on time machines, analogy, or mere speculation and anecdotes. You just need to be cautious when using them. The fact that scientists have used the same methods we used in the Nature paper for decades of studying living animals gives me strong confi! de! nce in the applicability of these modeling methods to ext! inct anim als, particularly when carefully approached with sensitivity analysis." I suppose the anatomist could simply respond by saying "As for sensitivity analysis, of course you are going to know the limits for the ostrich... Go and dissect one to find out... Go and dissect a few more while you are at it... Hell, go to town on extant every animal you can find. Be sure of yourself. Study those soft tissues. Make sure all those equations work for body forms that exist today. But excuse me... when it comes to extinct animals for which no exact extant analogue exists to test your numerical model against, why are you so sure that you have accounted for all of that extinct animalâs anatomical specializations and their importance? You can freely go ahead and say that this is turning into speculations and anecdotes... but exactly how are you going to dissect a tyrannosaur to see if you included all of what Mr. Hutchinson calls "strings and sealing wax and other fancy stuff"?" This is al! l ! answered by Occamâs Razor right? But what did I say above when it came to simplifying the goods when it came to numerical weather models? And by the way... when it comes to the weather, numerical models are not a very good supplement to other lines of evidence.... Looking out the window is. The anatomist could also cite examples of failed numerical models with the talk of diving whales, swimming tuna, jumping frogs, and running turkeys. Numerical modeling works best if you have solid foundations for its premises... but not so great for trying to deduce the premises in the first place. Who is the one with tunnel vision again? I forgot. I tend to think that my thoughts can be pushed aside as Mr. Paulâs were... The consequences of not understanding how this particular numerical model works. And for me, this is probably true. Being that I cannot articulate what I want to say as good as Mr. Paul does, doesn't help me either. But in the same breath, I suppose I could also pick up the mantle of BAND and Mr. Feduccia to proclaim that Dinosaurologists think that birds are dinosaurs because they donât understand avian anatomy and digit development and evolution. I suppose I could also set a near-unreachable criteria for usurping my conclusions... something akin to asking for a trackway of a running "Tyrannosaurus rex" would be suffice. Kris

- Prev by Date:
**Re: Spinosaurid Synonymy** - Next by Date:
**Re: Numerical Models and Running Rexes** - Previous by thread:
**Re: What is biomechanics? (or, The Truth About Flying Snakes - Was: Re: science** - Next by thread:
**Re: Numerical Models and Running Rexes** - Indexes: