This paper estimates the radiative fluxes and heating rates of absorbing aerosols, scattering aerosols, and cloud ice particles in tropical stratosphere using the newly developed radiative kernels. A notable merit of this study is the construction of aerosol kernels, and the application of them has the potential to better understand the radiative effects of aerosols in the upper troposphere and stratosphere.

Major comments:

Line 221. It seems that the kernels are calculated by perturbing aerosols at each level simultaneously. To my knowledge, previous studies all established kernels (e.g., water vapor, cloud) by perturbing the variable at each level at one time. It is necessary to clarify and justify the choice.

Line 285. The authors select four reference state boundary conditions for the kernel calculations. It is not clear why these four points are selected and why these points can represent clear-sky, low cloud, middle cloud, and high cloud conditions.

Line 298. It seems that the changes of water vapor and ozone in the upper troposphere and stratosphere are ignored in the simulations. Any estimate of its impact?

Line 299. The new kernels are constructed based on several linearity assumptions. However, it is not clear whether the cloud radiative effect varies linearly within a range of COD and whether the total radiative effects of aerosols and cloud ice can be represented as a linear sum of radiative effects associated with AOD and COD. It is necessary to validate these assumptions.

Figures 2, 3, 5-7, S3. The authors select several months (i.e., January, May, July) to validate the assumptions of kernel calculations. Are these months representative? Are the test results in other months consistent with the results in these months?

Other comments:

Line 200. Please correct the time range of reference state.

Line 314. “Figure 5 and Fig. 6”. Please use the uniform expression.

Lines 413, 420, 478, 496. Please correct the superscript and subscript.

Table 2. Please add RMSE for the kernel calculation.

During the Asian summer monsoon, strong deep convection helps lift aerosols from pollution into the tropical upper troposphere, lower stratosphere where it can impact radiation directly or by changing the thermodynamic and dynamic conditions of those layers.  This paper introduces new radiative kernels designed specifically to quantify these components of radiation change in the tropical UTLS where this Asian Tropopause Aerosol Layer (ATAL) occurs.  The paper shows these kernels are computationally efficient yet still able to properly represent the total aerosol radiative effects originally simulated directly by fully complex models.  This manuscript is well written, and provides a nice approach to constructing aerosol kernels, which is a much needed tool.  I provide some minor comments below.

Line 48: Matus et al 2019 also developed aerosol kernels that the authors may consider citing and discussing in the into: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019GL083656

Line 98-100: I appreciate the need to use different data sources for different variables, but did the authors evalute these combinations for inconsistencies?  For instance, are ERA5 temperatures sufficiently cold for instances where there is nonzero ice water content as reported by MLS? Or is the height reported by ERA5 a reasonable representation of where MERRA-2 thinks the aerosols are located? It would be useful to provide evidence of a sanity check to give the reader confidence that the dataset merging done here was appropriate.

Line 184-186: What is the shading effect of aerosols? And why is it not impacting the linearity of aerosol radiation at the tropopause? The authors should provide more discussion here about why linearity holds so well at the tropopause calculations.

Line 213-216: More explanation about why AAOD is so impactful on tropopause radiation and SAOD is not would be helpful here.  It may not be intuitive to most readers, who are likely more experienced with TOA or Surface conditions.

Line 258: I can appreciate that aerosol effects have low sensitivity to the background thermodynamic or cloud state? But what about sensitivity to the background aerosol state?  I would imagine a 10% aerosol perturbation in a high aerosol concentration condition vs a pristine condition would lead to a different radiative perturbation.  And likewise, the heterogeneous spatial pattern of aerosol base state would matter. Is that the case? If so, does it mean the kernels are only relevant for simulations where the background aerosol fields are similar to those of MERRA2?

Line 364: Mention of creating cloud kernels felt quite sudden as there was really no discussion of it in the intro.  There are mentions of aerosol-cloud interaction but I thought that was in reference to setting the base state for the radiative transfer calculation. I recommend the authors spend a little more time in the intro or in this section explaining the motivation for making these cloud kernels and why its a natural fit to make these particular cloud ice kernels along with the aerosol kernels

Conclusion: Especially since we are heading into CMIP7, I recommend a summary of the types of model various, levels and temporal resolutions one would need to apply these kernels to diagnose radiative effects.  I suspect subdaily data is always needed given the preservation of diurnal information in the kernels, and often modeling centers do not provide more than monthly or daily mean data.