This study quantifies four types of uncertainties in cloud radiative heating using an idealized extratropical cyclone simulated in a kilometer-scale model. It highlights the importance of improving ice-optical properties and cloud horizontal heterogeneity in numerical models, the two most contributing factors to the uncertainty. The experiment is well-designed, and the manuscript is well-written. I have a major question regarding the statement on the uncertainty due to the 3D cloud radiative effect, as well as a few minor questions for clarification. Therefore, I recommend a minor revision to this version of the manuscript.

The authors argue that the uncertainty contributed by the 3D cloud radiative effect is generally small. However, in Section 4, the local uncertainty from the 3D cloud radiative effect is not quantified. 3D cloud radiative effect has been shown to affect less on domain-average flux quantities but more on the flux distribution (i.e., the flux gradient at the cloud boundary). It is expected that the effect is more dominant in the local uncertainty. The authors can mention the Monte Carlo noise as a caveat when interpreting the results, but it is not the reason to exclude this term from the analysis. Otherwise, it is difficult to argue that this uncertainty is small compared to other terms.

Specific comments:

Figure 1: Does the domain change in the nine snapshots that are used for analysis? If not, does it indicate that the extratropical cyclone is stationary within the 4-hour timeframe?

Figure 5: I suggest adding another panel showing the net radiative heating difference between 3D and 1D calculations so as to be consistent with the following figures.

Line 263: “There is a direct relationship between cloud-side illumination and solar zenith angle.” Related to the question above, does the cloud field change significantly within the 4-hour timeframe?

Lines 273-275: “The stronger shortwave cloud-side illumination […] is most likely due to the higher solar zenith angle at higher altitudes […]” I think this argument can be verified from the flux output if the authors retain those outputs.

Figure 6: The time in the legend is confusing. I had thought it was local solar time, but the solar zenith angle is monotonically increasing from hours 10 to 14. I suggest explaining how these solar zenith angles are calculated or simply using the local solar time to avoid confusion.

Lines 359-362: I am confused about these sentences. I don’t understand why “the ice-optical parameterization has similar impacts on local and mean uncertainties” implies that “uncertainty due to the ice-optical parameterization is less important for km-scale circulations”? It sounds to me that the ice-optical parameterization is not necessary to be addressed in km-scale numerical models.

Line 411: The reference of Fan et al., 2022 is not up to date. Please use: Fan, C., Chen, Y.-H., Chen, X., Lin, W., Yang, P., & Huang, X. (2023). A refined understanding of the ice cloud longwave scattering effects in climate model. Journal of Advances in Modeling Earth Systems, 15, e2023MS003810. https://doi.org/10.1029/2023MS003810.

Technical corrections:

Currently, I found no correction needed.

This paper discusses sources of uncertainties in cloud radiative heating, which are crucial for understanding cyclone lifecycle forecasted by model. Authors investigated the uncertainty contributing factors of 3D cloud radiative effects, ice optical parameterization, and cloud horizontal and vertical inhomogeneity by employing 3D radiative transfer calculation based on Large Eddy Simulation cloud fields. The paper is well structured and written, provides valuable insights for bridging 3D radiative transfer with numerical forecast model, which I truly appreciated. I have some comments below and recommend publication once the comments are addressed.

General comments:

1. 3D cloud radiative effects versus cloud heterogeneity

Authors treats 3D cloud radiative effects and cloud heterogeneity (horizontal and vertical) as two distinguished uncertainty factors, but I think the terminology of 3D cloud radiative effects includes both. Horizontal photon transport (3D effects) occurs not only at clouds-to-clear-sky (what author refers to as 3D cloud radiative effects) but also at clouds-to-clouds (what author refers to as cloud heterogeneity). I would appreciate if authors can either 1) add more clarifications on why they sperate into two or 2) combine them into one.

1. Resolution effect

The resolution effect (modeling grid size in x, y, and z) is a quite complex factor as it can alter cloud fraction, cloud structure (e.g., vertical overlap), and cloud heterogeneity etc. Can authors provide some comments/insights on which plays the most important role when speaking of resolution effect to the CRH uncertainty (discussion from L391 to L402)?

1. Comments on “3D cloud radiative effects are overall small”

I don’t think with current results setup one can drew the conclusion of 3D cloud radiative effects are overall small not only because comment (1) but also because only limited solar geometries have been investigated. More importantly, since the paper only provides average profiles of CRH, the cloud 3D effects can potentially be “averaged out” (e.g., Figure 5c shows biases altering sign on the left and right of clouds). The realistic pattern of energetics (3D calculation) might play an important role in the convolution of cyclone, which cannot be captured by 1D even their averaged CRH seemingly the same. I would recommend adding standard deviation (or a selection of pixels) of the CRH profile in addition to the average value.

1. Offline 3D radiative transfer calculations

Although might be technical, if authors can provide some information on the computational time along with computational resources that have been used, either in the manuscript or in the response to reviewers, would be much appreciated.

Minor comments:

LibRadtran: I think should be “libRadtran” with lower case “L”.

P1L21: I would appreciate if authors can expand cooling and warming from the radiative perspectives of shortwave and longwave.

P2L28: It would be better to add some specific numbers (e.g., “~1km”, “~20km”) when describing process scale and synoptic scale.

P2L50: To me, cloud-side illumination effect is the same as cloud-side radiation leakage. Please elaborate on the difference.

P2L54: Suggest changing “might also lead to noticeable” to “can also lead to”.

P3L57: Please elaborate on “insufficient observations”. Aircraft in-situ observations have a decent amount of ice cloud observations for case studies.

P3L85: Please specify the vertical resolution after “75 model levels are used”.

P4L103: “homogeneous solver” to “1D radiative transfer solver”.

P5L124: “WBC” to “WCB”.

P6L147: “1.5 km” to “1km” (from readings on Figure 2b)

P8L204: What is the azimuthal direction? 0 at south (normally zero at north) and positive clockwise?

P9L205: Suggest changing “obtain low” to “reduce”.

P11L248, P11L251, and P18L376: “horizontal radiative transfer” to “horizontal photon transport” (or “horizontal radiation transport”).

P11L250: The cloud shadowing in Figure 5a indicates a solar zenith angle of 25° instead of 65°, please double check.

P11L254: “southern sides” is difficult to infer from Figure 5 only, please reference to long-lat plot of Figure 1.

P18L386: Rephrase “who showed that … showed that …”