La lune


Documentation > Greenhouse effect > Trying to guess the future > Should we trust climate models?

Should we trust climate models ?

last modified: august 2007

website of the author : - contact the author:


Let's go directly to the point : the answer is essentially yes. For those that do not accept a single-word answer (some people are SO defiant...) hare are a couple explanations:

there is not only one but 15 models in the world,

There is not just one country that can boast a model but any country that has a significant research activity is involved in climate simulation,

If a scientific team came to the conclusion, after having conducted serious simulation work (serious means reviewed by their peers, then published in a first rank scientific paper), that Earth will not "heat up" in response to human induced greenhouse gases emissions (a conclusion that is almost 2 centuries old, by the way), it would become world famous in a second (and would be a serious candidate for the Nobel Prize !), and one can be sure that many people tried very hard to achieve (properly of course) such a result.

One of the experiences performed to test the models if of course to check that they correctly account for what has really happened (below).

The above charts picture, for each region of the world and for the 1905-2005 period (and, at the bottom, for the whole world, the global land, and the global ocean):

with a black line, the decadal mean (or more exactely the difference between the decadal mean and the 1901-1950 mean) for the whole region, obtained from observations (dotted lines correspond to very little data available).

the pink zone corresponds to the enveloppe of 58 simulations performed with 14 different models when all forcings are taken into account (that is everything that happened is acounted for: volcanoes, solar irradiance, etc, AND human emissions of greenhouse gases),

the blue zone corresponds to the enveloppe of 19 simulations performed with 5 different models assuming all forcing are taken into account, except human emissions of greenhouse gases (a bit as if our emissions never happened).

Two major conclusions derive from this graph:

models correctly account for what really happened, thus backing the idea that they give reliable trends for the future (but those trends will always remain dependant on hypotheses on future emissions),

obviously no model can account for what happened since 1970 if we exclude human emissions : it is virtually certain that the warming observed since 1970 is the beginning of our footprint on the climate system.

Source: IPCC, Summary for Policymakers of the 4th assessment report (working group 1), 2007


The "trust index" is therefore very high concerning the forecast of a global warming as a result of anthropic (due to man) emissions.

The scientific community now considers that any person asserting that human greenhouse gases emissions are of no importance to the climate must demonstrate this affirmation.

As of today, no team (because modelling is team work) has published in a scientific paper results that would question the main qualitative conclusions of climate models (what does not prevent, from time to time, an individual who never published anything in a scientific paper to try to dismiss simulations in front of a camera or in a newspaper. So be it...).

A first indication on the relative fiability of models comes from the following operation: instead of studying the future, to know what will happen later, the model can be used to "study the past", which means that the run starts from 1860 instead of starting from 2000, in order to compare the output of the model with the instrumental data (1860 corresponds to the beginning of sufficient instrumental data to know what was happening on the surface of earth).

Even though the models do not reproduce exactely year by year values, they quite faithfully reproduce all the pluriannual trends, that is trends that last for a couple years or more.

There are other recent evolutions that match the results of simulations:

temperatures now rise faster during winter than during summer (what models foresee),

temperatures now rise faster during nightime than during daytime (what models also foresee),

a number of models have been tested on Mars and Venus, where they account properly for what is observed ; Venus has a climate system simpler than that of earth (no ocean, no ice caps, no biosphere) but includes a very specific phenomenum in its atmosphere (that rotates much faster than the planet itself) which is correctly reproduced by models.

on our good old planet, climate models also properly account for the location of the main climatic zones (no tropical climate in the middle of the south pole !), the rythm of the seasons, the pluriannual oscillations (El Nino, NAO...), the oceanic and atmospheric various currents (jet streams, trade winds, Gulf Stream...), etc.

The scientific community has therefore a very high degree of confidence in the climate models for their main qualitative conclusions, including the fact that earth will globally warm up in response to the additionnal greenhouse effect due to our industrial civilization. That does mean, of course, that everything is settled: an intense research activity still goes on to ameliorate the models so that they are able to reproduce more and more detailed evolutions.

But local comparisons (will the warming be more intense in Germany or in France, for example) are very difficult to establish, and will probably remain so....until things really happen ! Indeed, the climate system includes many complex equilibriums, and a small variation somewhere may have a significant effect later on and in a remote place.

Apart from a couple of indicators (average temperature, or average precipitations) on some large zones (a continent for example), it will never be possible to have detailed regional forecasts. This is a normal limit to climate simulation, because the system itself is not foreseeable on a detailed regional scale (impossible to give a forecast on what the temperature will be in Hong Kong the 6th of July, 2045 !), and this must not let us think that the general conclusions are flawed or that there is no danger.

Ruling out simulations because they are global by nature is alike considering that, if nobody can give it to the second, it's of no interest to know the approximative transportation time from Washington to Boston (or it's of no interest to know the approximative amount on a bank account if it can't be given to the cent, etc): some forecasts may be impossible to establish in detail, but are nevertheless reliable and useful for their main trends and large masses.


Back to climate change index
Back to the top