Maynard’s Guide to Global Warming

Maynard’s Guide to Global Warming
The problem is, the planet's climate is so complicated, you need a computer model to tell what's going to happen when you change a parameter.

The reasonable question you’ll ask is: How will we know the computer model is correct? Everybody’s building models. They’re all imperfect. How do we test a computer model?

The answer is logical. We test a computer model by feeding it old data, and evaluating its ability to “predict” events that have already happened. For example, we evaluate the data for 2005 to “predict” the climate for 2006. Then evaluate the data from 2006 and get a “prediction” for 2007. The valid model will prove itself by perfectly “predicting” the past. This will be rigorously tested and verified by the academic institutions of the world.

The model that can correctly “predict” the past can be expected to do well in predicting the future.

And the $64 question: How do the models do when tested this way?