Real Climate has introduced a weblog titled FAQ on climate models. There are quite a few issues that can be raised with their answers, but I will focus on just one here. It is their answer to the question “What is tuning”. They write
“What is tuning?
We are still a long way from being able to simulate the climate with a true first principles calculation. While many basic aspects of physics can be included (conservation of mass, energy etc.), many need to be approximated for reasons of efficiency or resolutions (i.e. the equations of motion need estimates of sub-gridscale turbulent effects, radiative transfer codes approximate the line-by-line calculations using band averaging), and still others are only known empirically (the formula for how fast clouds turn to rain for instance). With these approximations and empirical formulae, there is often a tunable parameter or two that can be varied in order to improve the match to whatever observations exist. Adjusting these values is described as tuning and falls into two categories. First, there is the tuning in a single formula in order for that formula to best match the observed values of that specific relationship. This happens most frequently when new parameterisations are being developed.
Secondly, there are tuning parameters that control aspects of the emergent system. Gravity wave drag parameters are not very constrained by data, and so are often tuned to improve the climatology of stratospheric zonal winds. The threshold relative humidity for making clouds is tuned often to get the most realistic cloud cover and global albedo. Surprisingly, there are very few of these (maybe a half dozen) that are used in adjusting the models to match the data. It is important to note that these exercises are done with the mean climate (including the seasonal cycle and some internal variability) – and once set they are kept fixed for any perturbation experiment.”
They make the following remarkable claims:
1. “With these approximations and empirical formulae, there is often a tunable parameter or two that can be varied in order to improve the match to whatever observations exist.”
First, there are always tunable parameters within each parameterization, and there are always quite a few more than one or two.
In my class on modeling, the students have documented the number of tunable parameter for a range of parameterizations, and 10 and more are common for each individual parameterization (e.g. see the class powerpoint presentations at ATOC 7500 for my most recent class).
Second, the only basic physics in the models are the pressure gradient force, advection and the acceleration due to gravity. These are the only physics in which there are no tunable coefficients. Climate models are engineering codes and not fundamental physics.
The framework of all climate models is illustrated in one of my powerpoint talks for weather models (see slides 3 and 4);
Pielke, R.A., Sr., 2003: The Limitations of Models and Observations. COMET Symposium on Planetary Boundary Layer Processes, Boulder, Colorado, September 12, 2003.
2. “Adjusting these values is described as tuning and falls into two categories. First, there is the tuning in a single formula in order for that formula to best match the observed values of that specific relationship. This happens most frequently when new parameterisations are being developed.
Secondly, there are tuning parameters that control aspects of the emergent system. “
The claim that there is a single formula is incorrect; the parameterizations of the physics involves subroutines (or their equivalent) which involve quite a few lines of code. More importantly, the matching with observations to tune the parameterizations is typicially completed for ideal situations (such as from a field observational campaign or high resolution model) and then applied to climate model situations which quite frequently fall outside of the conditions that were used to tune the parametrization.
Some parameters (such as the von Karman “constant”) are assumed to be universal, but most are just values that provide the best fit of a parametrization with the observed data used in its construction.
The second type of parametrization is the same as the first (their division into two types is artificial), except there is no observational data to make the tuning. Besides gravity wave drag, and a threshold of relative humidity for the onset of precipitation, a good example of a parameterization without any observational tuning is horizontal smoothing (which represents horizontal subgrid scale mixing).
The conclusion with respect to the Real Climate posting on “What is tuning” is that they inaccurately presented the actual limitations of parameterizations. They also did not accurately report that tuning involves many more tunable corefficients than they report.
Their sentence that
“Surprisingly, there are very few of these (maybe a half dozen) that are used in adjusting the models to match the data “
is incorrect. The students in each of my modeling classes (see for the classes for modeling and scroll to the bottom of each for the students’ class presentations where they decomposed parameterizations in order to quantify the number of tunable parameters) have documented the large number of tunable parameters within each of the parameterizations. There are no exceptions; all parameterizations involve a number of tunable parameters.
Real Climate is “surprised” that there are “maybe a half dozen” tunable parameters. They should have not been surprised but have looked in more depth to ascertain if their conclusion was correct (which they are not). Climate Science would be glad to post a guest weblog from Real Climate if they disagree with the Climate Science conclusions.
Readers on want an in-depth analysis of the number of parameters used in selected parameterizations in atmospheric modeling can view this in chapters 7 to 9 my book
Pielke, R.A., Sr., 2002: Mesoscale meteorological modeling. 2nd Edition, Academic Press, San Diego, CA, 676 pp