Mostly just rehashing something I've mentioned in the past, but this time in response to Mark Kleiman's point regarding climate change that "uncertainty actually argues for more precaution, not less. Current estimates could be wrong in either direction; that's what 'error band' means."
Yes he's right, but there's more. The error bands involve uncertainty about currently-quantifiable factors, but there are a number of as-yet unquantifiable factors that with best-case scenarios have no effect but under any other outcome will make things worse. I'm not aware of unquantified factors that work in the reverse direction. In other words, the "true" error bars should be skewed to show a greater chance that things will be worse rather than better than the predicted outcome. All the more reason to start reducing emissions, quickly.
This point doesn't get enough attention, IMHO.