How should we interpret the growing disagreement between observations and climate model projections in the first decades of the 21st century? What does this disagreement imply for the epistemology of climate models?
One issue that I want to raise is the implications of the disagreement between climate models and observations in the 21st century, as per Fig 11.25 from the AR5.
Panel b) indicates that the IPCC views the implications to be that some climate models have a CO2 sensitivity too high — they lower the black vertical bar (indicating the likely range from climate models) to account for this. And they add the ad hoc red stippled range, which has a slightly lower slope and lowered range that is consistent with the magnitude of the current model/obs discrepancy. The implication seems that the expected warming over the last decade is lost, but future warming will continue at the expected (albeit slightly lower) pace.
The existence of disagreement between climate model predictions and observations doesn’t provide any insight in itself to why the disagreement exists, i.e. which aspect(s) of the model are inadequate, owing to the epistemic opacity of knowledge codified in complex models.
For the sake of argument, lets assume (following the stadium wave and related arguments) that the pause continues at least into 2030′s.
Further, it is important to judge empirical adequacy of the model by accounting for observational noise. If the pause continues for 20 years (a period for which none of the climate models showed a pause in the presence of greenhouse warming), the climate models will have failed a fundamental test of empirical adequacy.
Does this mean climate models are ‘falsified’? Matt Briggs has a current post that is relevant, entitled Why Falsifiability, Though Flawed, Is Alluring.
You have it by now: if the predictions derived from a theory are probabilistic then the theory can never be falsified. This is so even if the predictions have very, very small probabilities. If the prediction (given the theory) is that X will only happen with probability &epsilon (for those less mathematically inclined, ε is as small as you like but always > 0);, and X happens, then the theory isnot falsified. Period. Practically false is (as I like to say) logically equivalent to practically a virgin.
With a larger ensemble, perhaps there would be ‘some’ simulations with a 20+ year pause.
If the pause does indeed continue for another 2+ decades, then this arguably means that the scenario of time evolution of the predictions, on timescales of 3+ decades, has been been falsified. This then brings us back to the ‘time of emergence‘ issue, i.e. whether the climate models are fit for the purpose of transient climate predictions, rather than merely equilibrium climate sensitivity.
If the climate models are not fit for the purpose of transient climate projections, and they are not fit for the purpose of simulating or projecting regional climate variability, what are they fit for?