I was recently reading Zach Holman's blogpost on time and remembering that I've had to work with CMIP5 output that used a 360 day calendar, rather than a 365 day or a Greogrian (or Proleptic Gregorian) one.
Do models that store output in this calendar run on a more physically accurate timestep and then somehow interpolate to a different calendar, or are they actually simulating on a 360 calendar? If they're actually using a 360 calendar, how does that work? Is incoming solar radiation simply parameterised in an interpolated way across the year for 360 days of regular duration? I can't imagine that the days are stretched out to account for being 1.5% fewer.
Does any of this have a substantial impact on the results?