Please see the attached file.

Review of the manuscript “The impact of coupled 3D radiative transfer on surface radiation and cumulus clouds over land" by Tijhuis et al., 2024

In this manuscript, the authors study the effects of 3D atmospheric radiation transfer (RT) on the surface’s energy budget and cloud’s properties for cumulus cloud fields over land.  They analyze a unique dataset of a dozen Large eddy simulations (LES) with online calculations of 3D RT in the solar spectrum. This is unique since all LES simulations use 1D RT to decrease computational load. Here the issue is solved by using GPU for enhanced calculations.

The novel dataset allows studying the effects of 3D radiation transfer on clouds’ dynamics and general properties like cloud cover, thickness and water density.

I think that the paper is suitable for publication in ACP, the analysis is complete and convincing, and the paper is clearly written. I believe that after a few clarifications the paper will be ready for submission, hence, I suggest a minor revision.

Major comments:

1. L.88: Can the authors explain why they chose a skin heat capacity of zero for the interactive surface and how realistic it is? To a none expert in the matter, it sounds like this could cause quick and unrealistic warming of the surface that can highly influence shallow convection.
2. L.161-166: This part of the paper is unclear. Please explain if aerosols affect the dynamics of the simulations. In short, how are the microphysical processes handled? Are aerosol radiative effects and horizontal variability coupled to in the simulations?
3. L.180-187: Please explain how cloud cover is defined. This is a tricky definition that can make comparisons between different datasets complicated (especially models to observations). Since the 1D and 3D don’t show much difference, I suggest showing the sensitivity of 3D to different definitions or choices of thresholds.
4. L.320: It is reasonable to assume that changes in cloud properties like cloud cover and optical thickness are most important for global surface radiation and scene albedo. The current version of Fig. 3 suggests that other processes influence cloud cover more than the radiative transfer scheme (RT). This raises the question of how important 3D effects are for surface or total energy budget.         Can the authors elaborate on this in the discussion and maybe even compare the bias caused by using 1D RT with other known biases and uncertainties in cloud or atmospheric modeling (like the choice of advection scheme, model resolution, microphysical scheme, etc.,)?

Minor comments:

1. L.147: I suggest using different names for the decomposed effects. Uncoupled is quite confusing and at the start can also be interpreted by the reader as 3D-1D_rad3D. I would suggest something like Radiative-only. The cloud effect  could be referred to as the 3D-coupling effect or dynamic effects.
2. L.198: Can the authors please explain why they chose to use the characteristic length scale and what is its physical meaning? Why wasn’t a simpler measure of cloud size like mean size used?
3. L.209: I wonder what are the effects of these findings on the scene albedo (top of the atmosphere upwelling fluxes). If cloud cover is the same but the clouds are thicker, does it mean a larger cloud radiative effect?
4. L.224: Please explain what is the displacement distance. Does it change with the radiative transfer scheme? If LWP is higher then clouds might live longer and be more advected.
5. L.241: Can the authors explain how the spread is quantified? Since the presentation is of only 3 cases, statistical measures are ambiguous, could be better to simply plot all three time series.
6. L.149-255: It took me a minute to understand the discussion about the splitting methods. Might be clearer to mention the two methods by referring to Fig.1 or showing it mathematically (e.g., 3D-1D_rad3D vs. 3D_rad1D-1D).  
7. L.322-326: I think that the authors can show the role of 3D radiative transfer on Earth’s energy budget with not a lot of extra effort. What are the changes in atmospheric heating and top of the atmosphere fluxes? Does decreased diffused radiation on the surface means increased heating rates in the atmospheric or higher scene albedo at the top-of-the-atmosphere? Might be worth to have even a short discussion on this as well.
8. L.336: Is there a reason to assume that the findings of this paper will be different away from the mid-latitudes? Dror et al., (IEEE, 2020) showed that a dominant subset of such clouds doesn’t have a strong latitudinal dependence.

Technical comments:

1. L.17: In cloud and weather modeling communities Cloud resolving models are usually referred to course resolution models on a scale of 1 km.
2. L.90: Worth mentioning that RRTM is 1D, and explain, even in short, the ray tracing concept and the novelty of the GPU usage (in Veerman et al., 2022 line 90).
3. L.160: This is not very clear, does aerosol vertical profile change with time in simulations?
4. 5: Adding y-axis labels as in Fig.6 would make the figure clearer.
5. 6 captions: Which dataset is presented, worth mentioning it’s for all 12 days.