snowpack

If one of your major model inputs is off by up to 50%, what does that do to the precision of your predictions about degrees of temperature change a century from now?