In Judy Curry’s post at Climate Etc
she listed information from Knutti 2008 regarding why there should be confidence in the multi-decadal global climate models. I have reproduced below this text from Judy’s post.
Knutti 2008 describes the reasons for having confidence in climate models as follows:
- Models are based on physical principles such as conservation of energy, mass and angular momentum.
- Model results are consistent with our understanding of processes based on simpler models, conceptual or theoretical frameworks.
- Models reproduce the mean state and variability in many variables reasonably well, and continue to improve in simulating smaller-scale features
- Models reproduce observed global trends and patterns in many variables.
- Models are tested on case studies such as volcanic eruptions and more distant past climate states
- Multiple models agree on large scales, which is implicitly or explicitly interpreted as increasing our confidence
- Projections from newer models are consistent with older ones (e.g. for temperature patterns and trends), indicating a certain robustness.
I will discuss each of these criteria below
1. “Models are based on physical principles such as conservation of energy, mass and angular momentum.”
Models actually include basic physics for only a subset of the physics, This basic physics includes the pressure gradient forces (e.g. in the atmosphere; oceans), gravity, and advection (e.g. by winds, currents, percolation of water into the soil) on the resolvable scale of the model (which is at least 4 grid increments as I discuss in detail in Pielke (2002)). All other aspects of the physics, chemistry and biology are parameterized using tunable coefficients and functions. The multi-decadal global climate models (and indeed all numerical climate models) require the conservation of energy, mass and angular momentum, but the implication from the first bullet of Knutti 2008 that the climate models are basic physics code is incorrect.
2. “Model results are consistent with our understanding of processes based on simpler models, conceptual or theoretical frameworks.”
This is certainly a necessary test of any complex code. However, the pertinent question is are the model results consistent with real-world observations? For time periods decades into the future, there is no way to test this requirement. In fact, even in hindcasts of past years, the multi-decadal climate models have no regional skill, as I posted on in When Is A Model a Good Model?
3. “Models reproduce the mean state and variability in many variables reasonably well, and continue to improve in simulating smaller-scale features” and “Models reproduce observed global trends and patterns in many variables”.
This are erroneous claims. As shown, for example, in
Koutsoyiannis, D., A. Efstratiadis, N. Mamassis, and A. Christofides, 2008: On the credibility of climate predictions, Hydrological Sciences Journal, 53 (4), 671-684
where, among their conclusions, they write with respect to the global climate models that they
“…perform poorly, even at a climatic (30-year) scale. Thus local model projections cannot be credible, whereas a common argument that models can perform better at larger spatial scales is unsupported.”
Even Kevin Trenberth has written with respect to these models (see)
“…the science is not done because we do not have reliable or regional predictions of climate.”
4. “ Models are tested on case studies such as volcanic eruptions and more distant past climate states”
The testing of global climate models when there is a volcanic eruption is a much simpler evaluation than multi-decadal global model predictions associated with natural variability and the diverse range of human inputs into the climate system.
Large volcanic eruptions result in the insertion of large quantities of ash into the stratosphere which reduces the solar irradiance that reaches the surface. This produces a cooling with the spatial distribution of the cooling dependent on where the volcanic emission into the stratosphere occurs. This use of the global climate models is an effective test of its skill and prediction of global and regional climate on the time period of seasons to a couple of years after a major eruption, as real world data can be used to directly compare with the model forecasts. This is a valuable necessary (but not sufficient) test of the skill of global climate models.
The model simulation of distant past climates is much more difficult since observational verification of skill must depend on proxy data. As a result temporal and spatial resolution is coarse, and only the larger climate perturbations can be resolved (not tenths of a degree in a global average temperature, for example). Moreover, skill with these models with respect to proxy data often occurs primarily due to the imposition of the different topography of earlier times as the bottom boundary condition. With respect to the last glacial maximum, for example, the insertion of continental ice sheets that are thousands of meters high over vast areas directly alters the wind circulations in its vicinity such that finding proxies of cold vegetation on their boundaries is expected even without a model simulation.
The use of the global models, when there are major volcanic eruptions and for past climates, is a worthwhile scientific endeavor. However, it does not indicate if the models necessarily have skill with respect to predicting climate decades from now associated with changes in atmospheric greenhouse gas concentrations, land use/land cover change and other human and natural climate forcings.
5. “Multiple models agree on large scales, which is implicitly or explicitly interpreted as increasing our confidence” and “Projections from newer models are consistent with older ones (e.g. for temperature patterns and trends), indicating a certain robustness”.
Model to model comparisons, while interesting and necessary, are no substitute for comparisons with real world data. The models themselves are actually quite similar to each other in terms of their dynamical core (i.e. the pressure gradient force, advection) and their parameterizations of the physics. They are not independent tests of skill, as they are themselves hypotheses (expressed in a mathematical set of numerical code).
Finally, the Knutti 2008 list is remarkably silent on what should be the most important test of the multi-decadal global climate models. This test is
What is their quantitative skill at predicting climate variations and change on short (e.g. days); medium (e.g. seasons) and long (e.g. multi-decadal) time scales?
Until this test is completed (the “seamless climate prediction“), policymakers should not have confidence in their forecasts (projections) decades into the future.
For further relevant posts on this subject see