[cm2187]

That's the main problem I have with climate science. They don't really have the luxury of proving by experimentation, and I have seen enough perfectly good backtestings in my industry (finance) that go wildely off as soon as they go live, that I have a fundamental mistrust of backtesting as a way to calibrate a predictive model.

[DennisP]

From the article: "A recent study compared models as old as 1970 with observations made in the decades since. Some models warmed up too much, and some too little, but 14 of 17 past projections turned out to be consistent with the measured path of global average temperatures."

So we actually did an experimental test of the predictions made in 1970, and those predictions turned out to be largely correct.

We should probably halt this experiment on ethical grounds.

[New_Horizon]

Especially because they seem to test it by making it predict history, but we don't have anywhere near the same amount of historical data. So what are you testing, really? Just that it accurately finds average temps?

[bannable]

Climate models as far back as the 1970s have predicted the temperature changes in the time from then to now -- those are not predictions that are coming from hindsight, but rather predictions where their effects were observed over the last 50 years.

If you are unfamiliar with climate sensitivity, this is a good place to start: https://skepticalscience.com/climate-sensitivity.htm