Why this claim is wrong: Here's a recent
study of climate model accuracies. In 2007
a study by Douglass et al. was claimed to show that models
were overpredicting the recent past warming when fed with past data, apparently
casting doubt on climate models. But it has been shown that the Douglass paper did
a poor job of accounting for systematic errors in the observations and an
extremely poor job of calculating error estimates on their conclusions (which
is key if statistical significance is to be claimed). The full
discussion is a bit technical, but the adventurous can dive in. A more careful analysis was done by Thorne et al. (2007) at the same time and finds no statistical significance to differences between
model trends vs observational trends in tropical
temperatures. As for "kluge's",
this word is designed to make you think climate scientists are simply assuming
what they are already convinced is true. No. All computer models of some phenomena must include simplified formulae for certain physics in order to be run in a finite amount of time on real computers. Modellers verify that the simplified formulae give reliable results by a variety of tests before doing "production runs" for their research. Tests such as using past data to generate forward predictions which in fact are already in the past, and verifying that the "post-dictions" are good. Even simplified formulae for physical effects need to be justified in a paper by appeal to real physics. One simply can't make up a bogus equation which violates physical principles and expect the journal referees to accept it. Another standard test is to change the resolution of the modelling grid, both spatial resolution and time resolution, and see if the results are robust (i.e. they don't significantly change). This gives confidence that your predictions are not far off due to finite numerical resolution. It's obviously never going to be possible
to calculate the future behavior of a system of a quintillion quintillion quintillion atoms using only Four Fundamental Forces of Nature. It doesn't have to be. Critics will try and claim that the "butterfly effect" in chaos means that even a tiny imperfection will completely destroy your prediction. Such critics are naive, and don't understand the difference between weather and climate. We may never be able to predict whether it will rain in Moscow exactly 2 years from today, but we can make statistically averaged predictions, which are strikingly accurate. Here's an analogy - imagine getting your nose right up to the water of a rushing mountain stream and try and predict exactly what bubbles and eddies will be in front of your nose in 1 minute. It's impossible. That's weather. But get 100 yards away and ask what the stream will look like in 1 minute and you'll be able to get a pretty accurate picture. Models
have improved significantly since the 1990's, and continue to give strong support to the conclusion
that global warming is primarily due to CO2 from fossil fuels, with secondary
add-on damage from human-induced methane, deforestation, and certain types
of air pollution. Here is a FAQ on climate models. While this
9 minute YouTube has no details,
it does show good visuals on how climate models compare to observations over
the years, with a bit of history. This video shows how numerical resolution has improved dramatically over the years, while the fundamental conclusions remain the same.
Stanford University's Professor Ken Caldiera points out here, in a recent interview, that climate models even in the early 1980's were predicting all of the following human-caused climate change effects: That human-generated greenhouse warming would assert dominance of the changing global average temperatures by the 1990's, that the troposphere would heat up while the stratosphere would cool (only greenhouse warming can cause this), the the oceans would soak up most of the heating, the the northern hemisphere would warm up significantly more than the southern hemisphere, that the poles would heat up far more than the rest of the globe. All of these have proven quite accurately true by subsequent data from ~1990 onward.
Global warming is real, it is caused by humans.
| In Short: Checks and balances on the realism of the modelling come at all stages, and a thorough error analysis must be done before any paper gets past a peer-review and makes it into a genuine scientific journal. Other researchers using other modelling codes and assumptions do the same, and differences are either resolved, and/or become a measure of quantified remaining uncertainties in predictions. The climate models of today have gone through these evolutions, and show excellent agreement with observations when using past data. Even climate models of the 1980's were predicting the key features of global warming caused by human-generated CO2 which were confirmed by the observations in the 1990's and beyond. The implication that climate modellers don't know what they're doing, is simply a lie. |