The advent of climate models has given an aura of certainty to climate predictions but a model is no better than the science behind it or the data that feeds it. Unless the weights and interrelationships of climate drivers have a solid scientific basis, feeding future predicted data into climate models lack the same. The same goes for data input. Models do not make the science. The models model the underlying science.
A climate model can be viewed as basically a function. The output value (temperature) is a function of the various values given to the chosen climate drivers (inputs) and their inter-relationships. Also a slight increase in the weight given to solar activity (or other drivers) with corresponding changes in weights given other input values can radically change future predictions while still modeling historical data (amusingly, researchers at MIT fed random data to the Mann hockey stick model and out popped the hockey stick!). The major natural climate drivers generally include orbital variations, solar output, volcanic activity and plate tectonics. Obviously, future volcanic activity cannot be determined but the same can be said for sunspot activity even though the Maunder, Sporer, and Dalton minimums corresponded with lower than average European temperatures.
There are, then, basically three problems that climate models face.
1. The internal weighting and internal relationships of climate drivers.
Climate Models May Overstate Clouds’ Cooling Power, Research Says — NYT
(7/4/16)
2. Ascertaining the values of current drivers
Different drivers, such as air pollution and volcanic eruptions, cause radiative forcing. A positive forcing has a warming effect on the planet, while a negative forcing cools it. The concept is straightforward, but it’s very difficult to establish the actual value of each factor that affects Earth’s energy balance. That’s because each is challenging to measure, and some of them overlap.
3. Dealing with projected values of drivers with no known periodicity.
Volcanic eruptions; cloud cover; CO2 absorption by oceans; Drying forests, methane bubbling up from Arctic wetlands, El Nino (a “breakthrough” study moves El Nino predictability from 6 months to a year); sunspot activity (the Maunder, Sporer, and Dalton minimums corresponded with lower than average European temperatures.
The third difficulty is freely admitted by government agencies like IPCC and NOAA. They make a distinction between prediction and projection.
The IPCC provides temperature “projections” as part of their assessment reports. They say they are not “predictions” since they are based on various scenarios involving different amounts of CO2 and other gases in the future.
The projections are of the form that, other factors remaining constant, if factor X is increased then thus and so will be the case. It is an admission that prediction is iffy. It is also worth noting that weather science may be pretty much “settled” but that doesn’t mean its predictions are thereby very accurate.