Ogmius Exchange Part II
Models can be useful tools for planning ahead: A response to Thomas Chase by Kevin E. Trenberth
The opening question raised is “Should policy makers base decisions on the results of current climate models?” Of course the answer is no. George Box is credited with saying “All models are wrong, some are useful”. It applies to climate models especially well. No one should base a decision on a climate model and its output without proper evaluation as to whether it is in the useful category. In fact models are used to guide decisions every day: weather forecasts, seasonal forecasts, and so on. But they should not be used as a “black box”.
A climate model is a tool; often a very sophisticated tool that encapsulates much of our understanding about the complex climate system. But it is still a model that makes assumptions and approximations, and is a grossly simplified version of the real world. Faster computers that can permit much higher resolution are required, for one thing, to merely capture our current understanding about the role of currently unresolved phenomena such as hurricanes. Adding more processes and complexity could also allow progress to be made.
Chase decides to use as a metric the radiation at top of atmosphere (TOA). Yes models contain biases and errors. But do they matter? The main way models have been used is to examine the change in response to some new forcing. This avoids worrying about specifying the initial state, and no model, even if perfect, would be expected to closely match the Earth Reduction Budget Experiment (ERBE) values for a very limited period (especially regionally) unless it were initialized and underwent the same sequence of El Niño and La Niña events. Models would also differ from each other depending on sequencing of such natural variability. Chase’s interpretation of Figure 1 is very flawed by not accounting for such effects. Models differ: e.g., in resolution, in land-ocean definition, in vegetation specification, and in basic things like the total solar irradiance. Rms errors in Figure 1 say nothing about the errors in the zonal or global mean. Errors in such quantities also say nothing about what happens when the climate is perturbed. The response to some forcing is what matters. If the response is linear or small, then the bias matters not a bit.
Chase should not mistake the uncertainty in knowledge about the forcing and how it has changed with the uncertainty in the model formulation. Aerosol forcing is poorly known. That uncertainty does not affect the confidence in the response to specified known forcing. He also mistakes forcings and feedbacks. Processes internal to the climate system, such as those involving the hydrological cycle, are feedbacks and properly do not belong in Figure 2. However, the hydrological cycle is one area where we suspect that nonlinearities matter. Nonetheless, the dominant feedback effect, that of increasing water vapor with increased heating, and thus an enhanced greenhouse effect, is simulated quite well by models and in ways consistent with observations. Uncertainties in clouds, and aerosol effects on clouds, however, remain large.
All of this does not mean that current climate models are not useful, though, in helping to guide policy decisions, provided they are used appropriately, with adequate evaluations of what they do well and what they do not, what their limitations are and what their capabilities are. This assessment is done by the IPCC.
The climate is changing and the past is no longer a good guide to the future. So what should we use for guidance? Any decision involves a model: whether it is a model of no change (which is surely wrong), a back-of-the-envelope or heuristic model perhaps based on someone’s limited experience, a simple energy balance model, or a full blown global climate model that requires a super computer to run. At least the latter includes many of the feedbacks and nonlinearities that we know are so important. But it does not include them all. Such models can be exceedingly useful if used wisely. Observed climate changes are now sufficiently large, and models in IPCC have now improved to the point that they simulate many of the observed changes going on. A confidence booster for sure! But we also need to improve models and have access to faster bigger computers.
Kevin E. Trenberth
National Center for Atmospheric Research
Continue to read Part III: Reflecting on good and useful climate models by Mike Hulme