By Willis Eschenbach – Re-Blogged From WUWT
It has been pointed out that while many of the global climate models (GCMs) are not all that good at forecasting future climate, they all do quite well at hindcasting the 20th-century global temperature anomaly [edited for clarity – w.]. Curious, that.
So I was interested in a paper from August of this year entitled The energy balance over land and oceans: An assessment based on direct observations and CMIP5 climate models. You’ll have to use SciHub using the DOI to get the full paper.
What they did in the paper is to compare some actual measurements of the energy balance, over both the land and the ocean, with the results of 43 climate models for the same locations. They used the models from the Fifth Climate Model Intercomparison Project (CMIP5).
They compared models to observations regarding a suite of variables such as downwelling sunlight at the surface, reflected sunlight at the top of the atmosphere (TOA), upwelling TOA thermal (longwave) radiation, and a number of others.
Out of all of these, I thought that one of the most important ones would be the downwelling sunlight at the surface. I say that because it is obvious to us—sunny days are warmer than cloudy days. So if we want to understand the temperature, one of the first places to start is the downwelling solar energy at the surface. Downwelling sunlight also is important because we have actual ground-truth observations at a number of sites around the globe, so we can compare the models to reality.
But when I went to look at their results, I was astounded to find that there were large mean (average) errors in surface sunshine (modeled minus observed), with individual models ranging from about 24 W/m2 too much sunshine to 15 W/m2 too little sunshine. Here are the values:
Now, consider a few things about these results:
First, despite the average modeled downwelling sunshine at the surface varying by 40 W/m2 from model to model, all of these models do a workmanlike job of hindcasting past surface temperatures.
Next, the mean error across the models is 7.5 W/m2 … so on average, they assume far too much sunlight is hitting the surface.
Next, this is only one of many radiation values shown in the study … and all of them have large errors.
Next, results at individual locations are often wildly wrong, and …
Finally, we are using these models, with mean errors from -15 W/m2 to +23 W/m2, in a quixotic attempt to diagnose and understand a global radiation imbalance which is claimed to be less than one single solitary watt per square metre (1 W/m2), and to diagnose and understand a claimed trend in TOA downwelling radiation of a third to half of a W/m2 per decade …
I leave it to the reader to consider and discuss the implications of all of that. One thing is obvious. Since they can all hindcast quite well, this means that they must have counteracting errors that are canceling each other out.
And on my planet, getting the right answer for the wrong reasons is … well … scary.