Dear Authors:
I have now reviewed the authors' comments and am happy to recommend publication after some minor changes are incorporated.
Best,
Graham Feingold


Main comment:

There is a tendency in this paper to take coarsely aggregated data and then explain them via detailed cloud process.

There is a broader question of the causal relationship vs. the general association between LWP and N_d. The associations are implicitly assumed to be causal when you resort to physical explanations as opposed, e.g., to co-variability of N_a and meteorological drivers that change LWP. This may fit into E4. At the very least, please point this out.


Abstract
check for grammar:
aerosol impact *on* LWP
Lines 9 and 10 : “in the relationship” repeated

Pg 3, line 25: Aren’t tau_c and r_e biases correlated? E.g., in broken cloud fields, r_e is biased high and tau_c biased low?

Pg 4, line 14: N_d (not N_D)

Pg 6: Doesn’t the fact that you give equal weight to all N_d bins bias interpretation? If an N_d doesn’t occur frequently then why give it similar weight? (Consider the co-variability comment above.)

Pg 8, line 5: You have done a nice job differentiating between the evaporation-entrainment feedback and the sedimentation-entrainment feedback, so why not refer to Xue et al. (2008) or Xue and Feingold 2006 here too.

Pg 9: line 32; see main comment. Here is one example. The Malavelle (2017) study shows that on average LWP is ~ constant. However there are large spatial differences across the domain. So I don’t see how you can invoke Malavelle here.

Pg 12, line 13, same comments as Pg 8, line 5.

Pg 13, line 7 and line 10: Have you actually shown that the meteorological conditions are similar across the years or do you surmise?

Pg 15, line 28 and Pg 16, line 5: I may not have been paying sufficient attention but I didn’t understand the 62%

Fig. 7: Is this dln albedo/dln AI or d albedo/dln AI?

Pg 17, line 18: “an” apparent

Pg 17, line 23: an early, and beautiful observational paper is Nicholls 1987? (QJRMS)

Pg 19: I may be wrong on this but I seem to recall doing this calculation and showing N_d dependent on sqrt(f_ad). Boers and Mitchell (1994) might have shown similar. In any case, I don’t remember it being linear. Worth a quick check.

Dear Colleagues,
I have reviewed your changes and am pleased to accept your manuscript for publication in ACP.
Best ,
Graham