This paper is, in part, a technical report of the updated infrastructure concerning the treatment of radiation in the Modular Earth Submodel System (MESSy), and in part, an evaluation of the performance of the newly implemented PSrad (Pincus and Stevens) radiation scheme vs. the ECHAM5 radiation scheme.

It is clearly written with sufficient technical detail to be useful for developers of the MESSy infrastructure as well as serving as a useful example for developers of other model radiation schemes.

The evaluation of the radiation schemes serves as a good test of the implementation and a useful evaluation of two schemes side-by-side in an identical model. The only problematic area is the comparison of the schemes against reference data presented in Pincus et al (2020), based on RFMIP (Radiative Forcing Model Intercomparison Project).

I would recommend this paper for publication once the following, generally minor comments have been addressed:

Principal comment:

1) Section 4, lines 630-640: The arguments presented here may be valid but it feels like the overall argument in this section is biased towards achieving a better comparison for the PSrad scheme. I think a more robust comparison could be done avoiding the need for the caveats in this section.

In the previous paragraph, lines 613-628, you use your present-day (PD) background runs to compare with the Pincus et al results for the forcing from pre-industrial to present-day GHG amounts. You scale the quantities to account for the different PD background conditions which sounds reasonable. For the CO2-folding experiments, however, you revert to the pre-industrial (PI) background runs. Your following arguments detail why this is a bad thing to do. Given that you have a range of CO2-folding experiments for the PD-background runs: CO2(pi), CO2(pd), 2xCO2(pd), 4xCO2(pd), you should be able to interpolate values for 2xCO2(pi) and 4xCO2(pi) to directly compare with Pincus et al. It would then be good to have all the Pincus et al results listed in table 7 to provide a clear comparison for the reader.

Minor comments:

1) Section 1, line 89: "resulted in 0.23 Wm-2": please define what this number represents, i.e. define radiative forcing as the difference in which fluxes? Top-of-atmosphere / tropopause / surface. Directionality?

2) Section 2.4 CLOUDOPT: Can you provide some details on how the cloud fractions are handled. Do you have separate ice and liquid cloud fractions or are they mixed in a single cloud fraction? How is the vertical overlap of cloud fraction handled? (Maybe a reference for this is sufficient.)

3) Section 2.5 ALBEDO, line 225: Please define what you mean by "blue-sky", "black-sky" and "white-sky" albedos. In other models, only the direct (your "black-sky" I think) and diffuse (your "white-sky") albedos are needed as the radiation scheme will solve for the direct and diffuse fluxes separately. Presumably the radiation schemes here don't do this and require a combined "blue-sky" albedo as well?

4) Section 2.5 ALBEDO: There is no mention of the spectral dependence of albedo. How is this handled by these schemes?

5) Section 2.5 Solar zenith angle dependent albedo, line 277: it would be good to explain at this point that you mean the fraction of diffuse and direct flux will be needed from a previous timestep call of the radiation scheme. What happens at model start-up when there is no previous call?

6) Section 2.6 (1): This appears to be an arbitrary functionality to add that could only degrade the physical accuracy of the results. Using the middle of the interval would appear to be the best of the options available. However, none of these options appear to consider what happens when the sun rises or sets during the radiation timestep. I believe the best approach (particularly for solar zenith angle) is to calculate the orbital parameters as a mean over the period of the timestep for which the sun is above the horizon. Was this considered?

7) Section 2.6 (2), lines 293-296: Not much point mentioning this adjustment unless you are going to explain how it was adjusted.

8) Section 3.1, line 340: It would be useful to give an approximate horizontal resolution in km for T42.

9) Section 3.1, line 357: "purely dynamic": I'm not sure what this means (in our usage, this would mean all the physics parametrisations are turned off, which is not the case here).

10) Section 3.2, paragraph at lines 433-444: I notice you specifically target clear-sky SW with albedo adjustments, but there is nothing to specifically target clear-sky LW. Is surface emissivity fixed for these schemes? Is there anything else that could be used to target this?

11) Section 4, line 550: Please explain how the stratospheric adjustment is done.

12) Section 4, line 619-620: "we assumed the 2014 values used by Pincus et al are similar to Meinshausn": I believe the values used by Pincus et al. are essentially those publicly available for RFMIP, so this assumption could be properly checked.

13) Section 4, line 628: the N2O RF presented by Pincus should be stated for comparison (even better, all the values from Pincus should be added to table 7).

Typos etc.:

1) line 11: "of sixth generation of the the" ->  "of the sixth generation of the"

2) line 55: "radiative RFs" -> "RFs"

3) line 86: "old radiation" -> "old radiation scheme"

4) line 351: table 2 is referenced before table 1

5) line 430: "adjust parameters target-oriented" -> "adjust parameters in a target-oriented manner"

6) line 679: "much increased (decreased) to the radiative forcings" -> "much increased (decreased) with respect to the radiative forcings"

This manuscript describes major updates to the radiation schemes within the Modular Earth Submodel System (MESSy), which is an infrastructure designed to link different submodels into the same framework to more seamlessly perform simulations with different model components.  Specifically, this work covers the implementation of the PSrad radiation scheme into MESSy, as well as updates to related submodels for calculating cloud optical properties (CLOUDOPT) and aerosol optical properties (AEROPT), as well as implementation into MESSy of a new albedo scheme (ALBEDO).  The authors find that implementation of these schemes leads to reduced biases in temperature and humidity of a handful of key climate processes and improvement in radiative forcing variables for greenhouse gases relative to reference values. I find it particularly valuable that the implementation allows for easier calculation of radiative forcing through online double calls. These calculations are important but not routinely performed at most modeling centers.  This manuscript is well written and will be of interest to GMD readers, especially as many modeling centers work towards updating their radiation schemes and, more generally, work towards stronger unification of submodels. I recommend some minor revisions detailed below.

General: I think readers would appreciate some information about computational performance when implementing the new radiative transfer scheme with more spectral bands. Was there a noticible increase in compute time with the new code and, if so, what steps did the developers take in an attempt to improve speeds?  

Line 206-207: It may be a bit surprising to some, me included, that the developers decided to add a secondary LW ice mass extinction option that comes from a model that is now a few generations old (ECHAM4).  What there a particularly reason to bring back this scheme? Some context here would be interesting.

Line 245-258: What is the role of this observational-based albedo climatology when the scheme is used to simulate climates beyond the present-day? Is the climatology used as a scaling factor to preserve seasonality? Is it only implemented for certain types of simulations?

Section 2.6-1: Some motivation for providing additional flexibility in the orbital offset would be helpful. The previous version, where the offset would always falls in the middle between radiation calls, seems like the most reasonable approach for any case.  Are there cases where another option is better? Some context would be helpful here.

Line 355: It is clear that the sets of simulations performed in this section have different radiation schemes (PSrad vs E5rad) but what about the modifications to the other relevant submodels discussed?  I suspect the simulations using of PSrad also include all of the updates discussed for CLOUDOPT, AEROPT, ALBEDO and the orbital offset. If so, this should be noted in the text or better incorporated into the experiment names for clarity.

General Section 3: The biases are presented clearly, and the authors focus on important ones, but I was hoping for some attempt to explain the causes of the bias, and particularly for situations where the e5rad and Psrad-driven simulation biases differ.  Establishing causation is difficult in many cases, but some general discussion or potential explanations from the authors would be useful here. Is the warm stratosphere bias from the PSrad simulation (compared to the cold bias from the e5rads) related to the new handling of the orbital parameter offset, for instance?

Also relevant to Figure 5: ERA5 has a known cold bias in stratospheric temperature from 2000 to 2006, The reanalysis was rerun for this period in a product called ERA5.1. I am unfamiliar with how large this bias was, but it would be interesting to see if the EMAC-PSrad bias is reduced for years outside of this range, or if ERA5.1 is used instead. Presumably the Figure 7 humidity bias is impacted too. Details here: 

https://confluence.ecmwf.int/pages/viewpage.action?pageId=181130838

Line 599-606: Is the reduction in methane RF from IRF for PSrad significant? A 0.01 W/m2 reduction from IRF seems quite small and may just be noise, especially when the reduction does not appear to be present for the pi simulation.  I mention this because although stratospheric adjustments related to SW absorption may be playing a role in a reduction, the Smith et al figure points to cloud adjustments playing in even larger role, an effect not being captured in this work. And recently, Allen et al. 2023 looked into the cooling from SW absorption of methane explicitly, finding much of it is driven by cloud adjustments, rather than a stratospheric adjustment: https://www.nature.com/articles/s41561-023-01144-z

Line 638-639: Yes, the Pincus pd background likely has a warmer surface thus CO2 forcing is stronger, but it also likely has a cooler stratosphere, which is arguably more impactful on CO2 forcing as highlighted by Jeevangee et al. 2021 and He et al. 2023.  Related, this may explain why the CO2 forcing from the PSrad simulation is smaller than the E5rad simulations. PSrad produces a warmer stratosphere and thus the CO2 forcing is smaller.

Jeevangee et al. 2021: https://doi.org/10.1175/JCLI-D-19-0756.1

He et al. 2023: https://www.science.org/doi/10.1126/science.abq6872