Yesterday I highlighted the inadequacies and assumptions in the scenarios which form the input for Global Climate Models (GCMs). Today, I'd like to pinpoint the inadequacies of the GCMs themselves. The key point to appreciate about most computer simulations in science is that they rarely simulate all the relevant processes involved in the system under study. For all but the simplest of systems, it is necessary to parameterise various things. The behaviour and effect of parameterised processes and objects is not simulated, merely set by hand. The parameters represent the average or approximate effect of the things which cannot be simulated.
Simulation methodology requires that one validates a simulation by testing that it is capable of reproducing known data as output. In the case of a GCM, validation requires the simulation to reproduce the known global temperature profile up to the present day for the known profile of CO2 and SO2 emissions. However, there is a methodological problem here which besets many scientific simulations: the setting of the simulation parameters (a process called calibration) has to occur at the same time as validation. One adjusts the values of the parameters until the simulation reproduces the desired output. There is an obvious problem here: if the underlying logic of the model is incorrect, that fact can be concealed during calibration by adjusting the simulation parameters until the combination of parameter values compensates for the inaccuracy of the underlying model, and the desired output is obtained. To reduce the probability of this, one needs to calibrate a simulation against a number of different scenarios; one needs to obtain a fixed combination of parameter values which are such that the simulation reproduces actual data across a range of different scenarios. It is less likely that a specific choice of parameter values could compensate for inadequacies in the simulation model across a range of different scenarios.
However, for GCMs, there is a big problem here: there is only one set of data to validate and calibrate against; GCMs can only be calibrated and validated against 20th century climate data. This means that GCMs cannot be properly calibrated and validated until the parameterised processes are de-parameterised, and properly simulated. In the case of GCMs, the parameterised processes are typically things like cloud cover formation, which have a significant effect on global temperature, but which lie below the resolution level of the simulations. Sure enough, the various GCMs used by the various climatological research institutions around the world produce a large range of different predictions for the global temperature profile in the 21st century.
There is an even more severe problem with the validation and calibration of GCMs against 20th century climate data: not one of the models is capable of accurately reproducing the temperature profile throughout the 20th century. The 20th century basically divides into three phases from a global climate point of view: there was a period of rapid warming, but negligible CO2 emissions between 1900 and 1940; a period of global cooling from 1940 until 1975 despite significant increases in CO2 emissions; and a second period of significant warming from 1975 until the end of the century, in concert with a significant increase in CO2 emissions. The GCMs are capable of reproducing the temperature profile from 1975 onwards, but they tend to struggle with the temperature behaviour prior to that. One can postulate that the cooling from 1940 to 1975 occurred because several volcanic eruptions pumped SO2 into the atmosphere, and this had a cooling effect. However, the rapid temperature increase in the first part of the century, without a corresponding increase in CO2 emissions, is difficult for the GCMs to explain. Perhaps the best explanation is that patterns of solar radiation, such as sunspot cycles, were responsible for this temperature increase. If so, however, then one must acknowledge that current variations in global temperatures are a function of both solar radiation and CO2 emissions.