Let's see.
1 The first test of a model is to see if it is stable. They can be started at, say, 10*C, and will (if well-designed) stabilise at 15*C (the correct temperature of Earth) and stay that way for thousands of (virtual) years unless the solar input or other forcings change.
2 Modern complex models will sometimes, unasked, generate jet streams, trade winds, depressions and anticyclones that would be difficult for even the most experienced forecaster to distinguish from real weather, and even major year-on-year variations like the El NiƱo–Southern Oscillation. (Physics world, registration required)
3 Computer models can be set running at, say, 1900 and can reasonably follow the observed temperature curve right up to the present time - "hindcasting". Early models had to have arbitrary correction factors built into them, but modern models do not need these. Contrarians make much of the earlier use of these correction factors which they call "fudge factors" or curve fitting, but this criticism no longer applies, or at least, not as far as I know.
4 Models can be tested against ice core data to see if they reproduce ice ages, working from the initial change (warming or cooling from orbital changes) to represent the observed temperature cycle. This they can do.
5 Models can be given data regarding the amount of aerosols thrown into the air by volcanoes, and can accurately reproduce the resulting temperature drop.
It is true that models still cannot reproduce the exact details of the temperature record. In particular, they do not reproduce the recent plateau in the temperature record since 1998.
Clearly, there is a fairly regular fluctuation in the temperature record which the models are not yet reproducing which is why models did not predict the present slowdown of warming. Is it due to not allowing enough amplitude for changes caused by the solar cycle?
Here is one of my infamous composite figures of observed temperature, models and sunspot cycle from 1980 to 2010:
The pink line is the temperature, the underlying black line is the computer generated temperature, the grey lines are the confidence limits, and the fuzzy line below is the sunspot cycle. There is a vague correlation up about 2000, but then the relationship breaks down.
Not particularly exciting.
Wonder if the PDO (Pacific Decadal Oscillation, La Nina El Nino cycle) is any help?
The green line is global temperatures, the little green cakes with a feather coming out are volcanoes (note how they cause the Earth to cool due to shading from their aerosols) and the bottom is ENSO, red being the warming el Ninos, blue being the cooling La Ninas.
And so it goes on. Models are not 100% perfect, but neither are they 100% wrong. They are servicable, and most important, they are improving all the time.
So can they project the future accurately? This will be the subject of a future post.
For more detailed info on climate models, go here
More posts about modelling on this blog:
Model-free climate sensitivity is still dangerous
"Data" does not necessarily prove the models wrong
Climate models - are they scientific and reliable?
Climare model projections - what mpact do ocean currents and solar variation have?
No comments:
Post a Comment