I use computer models for rocket design.
But at some point, if you have any sense, you build something and test it in the real world.
If you're doing computer models of climate, at some point you need to look out the window.
The 2006 ozone hole was a record, and 2008 was just short of it. The computer models now say that the hole won't begin "healing" until we're all safely dead, so I guess that's not a problem for the modelers.