re: #214 thedopefishlives

On the topic of AGW models, just how good are they getting to be these days? Last I checked, which was some time ago, the models had poor performance in predicting previous years’ data, which gave one a low sense of confidence in forward extrapolation. Is this still the case, and even if it is, is that even a valid criticism to make?

Great questions.

There are a couple of interesting points to be made about this.

1. Much of the model vs. Data controversy back in the day was caused because the main satellite used to collect stratospheric data had degraded in it’s orbit and hence was giving bad data. Once that was fixed, the models and the data matched much better. Also, as computers evolved and more power was available it was possible to do much finer detailed calculations.

Think of it as something like a taylor series. If you can only get the first order term, it isn’t that good (usually) at predicting the development of a system, though you can still say some overall and important things. If you get more an more terms though, you get better and better. The computers finally got big enough to add more and more terms. Now there is a deep, long and detailed discussion to be made about how good is good enough, or how many terms do you need before it is good enough. The short answer here is that the scientific community thinks we have it good enough.

Here is a pretty old page, but look at the graph of model output as put against hard data.

gfdl.noaa.gov

You will see that the simulation follows the data quite well.

2. The models continuously integrate new observation as correction terms in order to “keep them on track.” This makes their extrapolations even more accurate. We also keep adding more and more physics to them as we detail more and more small interactions. All of the models start with the big stuff so to speak. We are tweaking the fine details now. How those details get tweaked and put is, is also always based on direct observations to make sure we got that physics correct, before putting it in.