File talk:Climate Change Attribution.png

In the figure text a model is said to have an "ability" suggesting that it is close to the underlying true observation. This is wrong. A model is more or less always "right" in hindsight because it is tuned to show this (within its frames. To validate a model the projections should be compared with the outcome. At present very few models have a remote "ability" to do so. I have changed the text accordingly. /JPC Lindstrom