Dear Authors,
Thank you for taking the initiative to improve the manuscript. I can see clear improvements to the clarity of the analysis and statistic robustness. Based on your current responses, I cannot help notice a few things waiting to be better addressed in your revised manuscript:
1 Reviewers pointed out that the simulated sea ice is low at 400 ppm CO2 compared to published modeling studies. In your reference list, the Koenigk et al., (2013) paper suggested that for EC-Earth, September sea ice free threshold is around 500 ppm CO2. In RCP4.5 and 2.6, EC-Earth did not reach September sea ice free by year 2100. I tend to agree with reviewers, a comparison with present-day control, and explanation of the differences are needed to validate your PlioMIP2 run.
2. Reviewers pointed out potential numeric errors in the model integration. One review noticed that this error may lead to global changes in atmospheric lapse rate. This sounds quite alarming. Please examine the lapse rate and explain whether or not this relates to different results from EC-Earth present-day and RCP runs. I wanted to point out that ERA-interim or any reanalysis data are using models to fit data, the priority of reanalysis model system is not to conserve energy. Scientific literatures were published to alarm the community not to use reanalysis blindly to perform energy balance analysis. In general, you cannot compared the energy imbalance of reanalysis data to performance of a model. A model, by design, should conserve energy.
3. Reviewers pointed out that the time scale of interactions between ice and heat fluxes is below monthly. One reviewer is unsure about whether correlation at monthly scale can be used for causation arguments. Would it be possible to continue the run for another 30-yrs and run CFRAM code and spatial correlations at daily time scale? This will hopefully address these concerns.
4. The target time period of PlioMIP2 is mid-Piacenzian (at 3.205 Ma, belongs to the later part of Pliocene). Pliocene epoch spans 5.3 to 2.6 Ma with varying CO2, orbital parameters, and gateway configuration. The word “Pliocene” or even “mid-Pliocene” is inappropriate for discussing model results.

Thank you! I am looking forward to the revised manuscript!

Dear authors,

Thanks for taking time revising the paper. After two rounds of reviews, two out of three reviewers still hold strong opinion about the study. In particular, one reviewer questions of your application of CFRAM method and explanation of out of phase behavior in SST and SAT:
"The fundamental issue I have with this work is that the authors look at a 400ppm Pliocene run and conclude that the surface albedo feedback is stronger and hence contributes to Arctic amplification. This is not new. What is new here is the use of this CFRAM method, but it is not clear to me that the authors have applied this correctly over ice covered regions and open ocean regions. With regards to the seasonal cycle of SST/SAT, this can simply be explained by the fact that SST cannot get colder in winter than -1.8C, while the SAT does not get warmer than 0C in summer over ice. However, the SST can warm substantially in summer and the SAT can get much colder in winter. This by itself explains the out of phase behaviour. "

One other reviewer is unsatisfied with your explanation of linearity of the attribution method in particular the potential perturbation fluxes introduced by switching boundary conditions from PI to MPWP:
"Some of my questions on the linearity of the attribution method and the Pliocene experiment setup are still not well explained. I suggest the authors to add an uncertainty estimate to the attribution method. The differences in boundary conditions between the preindustrial run and the Pliocene run will introduce perturbation fluxes not included in the attribution framework. The implication for current results should be considered."

I lean towards accepting your work eventually, however, you do need to address the above concerns, as well as other detailed comments/suggestions by all three reviewers published in the discussion forum. Thank you!

I am looking forward to your response. Hope you have a great holiday!
Ran

Line 17: "...response analysis method (CFRAM) and an approach similar to equilibrium feedback assessment..." I believe what you meant is that CFRAM is an approach similar to equilirbium feedback assessment. If so, please replace "and" with ",".