Re: OT: interesting global warming quote found elsewhwere

On Sat, 2 Aug 2008 00:00:31 -0700 (PDT), bill.sloman@xxxxxxxx wrote:

On Aug 2, 3:44 am, John Larkin
<jjlar...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
On Thu, 31 Jul 2008 21:05:21 -0700 (PDT), bill.slo...@xxxxxxxx wrote:
On Aug 1, 4:03 am, John Larkin
<jjlar...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
On Tue, 29 Jul 2008 19:10:55 -0700 (PDT), bill.slo...@xxxxxxxx wrote:
On Jul 30, 11:37 am, James Arthur <bogusabd...@xxxxxxxxxxx> wrote:
bill.slo...@xxxxxxxx wrote:
On Jul 29, 4:19 am, James Arthur wrote:
On Jul 26, 5:40 pm, bill.slo...@xxxxxxxx wrote:
On Jul 26, 1:31 pm, James Arthur wrote:

<snip>

And that work's already been done.

Yep. It is called agricultural science, and it monitors how summer
follows winter, and the rain falls in close enough to same place in
close enough to the same volume from year to year to let you plant
crops and harvest enough of them to make money out of your
predicitons.

Your chaotic weather systems are unpredictable, but they deliver
roughly the same sort of statistical distribution of weather from year
to year, and you can model them well enough as statistical
distributions.

Sure, there's a long-term trend to climate: it's getting warmer, and
has been for nearly 20,000 years:

http://www.museum.state.il.us/exhibits/ice_ages/images/laurentide.mpg

Shortly before the trend was for cooling.  But those models were
proved wrong. :-)

Sure. We should be headed back towards the next ice age by now, but
pre-industrial anthropogenic global warming seems to have over-ridden
the small orbital forcing

http://en.wikipedia.org/wiki/Milankovitch_cycles

if William Ruddiman has got it right

http://en.wikipedia.org/wiki/William_Ruddiman

Your modeller's efforts freeze out the atmosphere and melt lead in the
same way that my LTSpice circuits occasionally produce megavolts and
mega-amps. This doesn't stop more carefully constructed models from
providing useful information.

You don't get it.

On the contrary, I get it rather better than you do. Like your
informant, I too have worked on computer models (albeit at a pretty
primitive level, back in the 1960's as part of my Ph.D. project). The
instability your modeller complained about isn't a fault of the model
as a model of the climate, but a fault of all models that rely on
numerical integration to step forward in time - which is why it shows
up in LTSpice models as well as climate models.

LTspice, or any well-written sim program, with properly selected time
steps, modeling most circuits, is dead stable to ppb accuracy. Change
a resistor value a little, and the sim time-series result changes a
little, just like real life.

But if you model a chaotic circuit, which you can also do in Spice,
the results become extraordinarily sensitive to small component
variations, time step, algorithm choice, and fp errors. Changing
anything will *totally* change the results after a while.

Very true, but not what I was talking about. If you model a perfect
stable circuit, but give it ill-chosen starting conditions, the
numerical integration algorithm breaks down, and you get absurd
results.

What? It's not the initial conditions that make a model unstable,

Read what I wrote. It isn't the model that is unstable, it is the
numercial integration process. I can only assume that got forced into
rounding error and couldn't recover.

it's the system being modeled, or bog-stupid circa-1960 sim algorithms.

This happened a few months ago on a up-to-date copy of LT Spice
running on a machine with a 64-bit AMD processor.

Double-precision FP math isn't likely to "break down" unless you
really try to make it.

I wasn't trying, though hadn't bother to update an old set of starting
conditions.

Charge an RC to a billion volts and Spice the
discharge; it will look just fine. There are few electronic systems
that would simulate unstable just because you set the initial
conditions wrong; and those systems probably *do* have real-life
latchup states that should be investigated.

It's more likely that Spice misses the latch states of a real system
than that it invents them.

There weren't any latch states in the circuit - it was just
headingsmoothly off to infinity

Are you saying that a real circuit wouldn't behave this way, but
defects in LT Spice made it diverge? That Spice grossly inaccurately
modeled the behavior without some wrong setup? After all, you can make
a paralleled RC oscillate in simulation if you really want it to.

Please post the circuit.

Are you saying that the climate models are unrealistic too?

John

.

Loading