H. Sone and I actually discussed these main topics at work other day... greenhouse gasses and climate change. We talked about the undeniable facts, such as the increase in CO2 levels over the years and it's role as a greenhouse gas (transparent to blackbody radiation @ 5780 K, but opaque to radiant E @ ~300 K). However, we both expressed doubts about the long-term accuracy of computer modeling.
In fact, when I took a numerical methods course last year, the Professor stressed a healthy skepticism for ANY results spit out by a computer. I, personally, will trust climate modeling when we can predict the weekend weather with 100% accuracy.
My labmate talked of attending numerous environmental/earth science seminars on global warming. He said students there defended these sprawling models with "the computers are much more powerful now".
Well, yeah... so now we can crunch larger volumes of uncertain numbers to a finer resolution.
Hooray!
The common problem with numerical tools is not the processor speed or memory (although they can be limiting), but faults in the underlying equations and constants. If the governing system is sketchy and the physical parameters (such as specific heat, density, etc...) are inaccurate, these slight gaffes will propagate through your model and return essentially garbage... especially since discretization inherently introduces errors from the very beginning.
But I guess that's the crux of an engineer-- many times reality presents far too many variables to effectively model and solve (I'm looking at YOU, turbulence). Not saying we shouldn't try; by all means, smart, simplified calculations can be extremely valuable and often point in the right direction.
This is just a gentle reminder that while all models are wrong, some are useful.
I guess I'll go ahead a stick a 'politik' label down here. Ugh.
No comments:
Post a Comment