This is an easy-to-read and useful introduction to NCAR’s perturbed parameter ensemble (PPE) methodology.  The results are interesting and important.  However, I think that some additional discussion by the authors could clarify the interpretation of the results.

Major comments:

Lines 77–78:  “we initially created 250 different sets of parameter values in addition to the default CESM2-CAM6 setup (total of 251 sets).”  How was this number chosen?  What is the consequence if only half of the 250 parameters sets are emulated?  The paper contains a nice comparison of emulator techniques, but I wonder if an excellent emulator could be thwarted by a sparse sample of parameter sets.

Line 241: “First, we will illustrate the basic spread across the PPE for several key features of the simulated climate system.”  Is this the ensemble spread over all values based on a uniform sample of parameter values within the expert-chosen ranges?  Could you provide some comments on how these spreads should be interpreted?  The magnitude of the spread would be expected to depend sensitively on the chosen parameter range.  If that range is chosen subjectively, then the spread will inherit that subjectivity.  In principle, the range might be problematic because, e.g., we know that some ensemble members produce an unrealistic climate, because the authors state on lines 409–410 that “First, we will illustrate the basic spread across the PPE for several key features of the simulated climate system”.  Given this, are the spreads realistic?  E.g., the max ice fall speed factor is 25 times the min value, suggesting that we don’t know the ice fall speed within an order of magnitude.  Is this true?  To cite another example, the accretion enhancing factor varies by a factor of 100.  Is this degree of uncertainty realistic?

Minor comments:

Lines 44-46: “in the current study we perturb 45 different parameters, which would require a minimum of 3.5·10^13 (2^45) simulations using OAT if each parameter was tested with only two values in all combinations.”  How are you defining “OAT” here?  2^45 sample points would fill the entire 45-dimensional volume of parameter space, which would involve perturbing all parameters simultaneously, not one at a time.  Perturbing each parameter individually would yield only 2*45 samples, no?

Table 1:  The max value of DCS is listed as 1.0e-6.  Should it be 1000e-6?

Line 247: “global, annual mean net top of atmosphere flux balance (TOA: Figure 3H).”  The definition of TOA is unclear to me.  Does a net downward flux have positive TOA?  Or negative TOA?  Is TOA different than RESTOM (line 297)?

Line 414: Replace “relative” with “relatively”.

Line 417:  Replace “Colombia” with “Venezuela”.