Very interesting paper, it would be good to see this study extended to other cases. I include a few minor comments in the pdf attached. 

This manuscript discusses the cause of the observed thermal infrared signal of a specific biomass burning episode in western US (the Dixie fires). Different causes are investigated using observations and radiative transfer modelling, and a conclusion is reached that the most important player is the surface cooling by the smoke plume. The data and analysis presented are mostly convincing, although not always straightforward to understand. I would appreciate seeing some additional physics in the discussion (examples given below in the specific comments), and I wonder why CrIS hyperspectral radiance data (in addition to its T and WV profiles, and skin T) is not examined together with the broad TIR channels of GOES and MODIS. That would show pretty nicely (at least it does in the IASI data I have looked at) how the TIR radiance is "flatly reduced" by about 25K for smoky observations, while the (low intensity) gas absorption lines are not toomuch affected. Such a flat reduction of atmospheric window TIR "baseline" BT directly points to a (skin) surface temperature reduction or a thin cloud (aerosols usually have specific signatures with slopes and/or different impacts on radiance at about 9 and 11µm - see for example Clarisse et al DOI 10.1364/AO.49.003713). The trick, I think, in this case, is to ensure that the observed lower BT is due to lower surface temperature and not to thin (water) clouds (ice clouds would have a typical TIR signature which is not observed). This is where combining the TIR with SWIR and visible observations comes in.

Specific comments and questions, suggestions:

- suggestion: mention the episode name in the paper title - it was important enough to even have a wikipedia page ;)

- lines 23-25: this sentence is rather unclear to me. I guess the authors means the very thick optical depth at visible wavelength? I am also very puzzled as to how a BT change in TIR could be mapped to a smoke signal, especially since the authors highlight that the smoke signal is due to surface cooling - I mean how would the difference be done between smoke and any other surface cooling reason, such as a cloud?

- lines 36-39: in addition to the scattering, the aerosol absorption plays a role in the TIR (usually a dominant role) and this can happen no matter the particle size - of course the particle number concentration needs to be higher for smaller particles to have an observable signal, and the particles need to be absorptive in the TIR

- section 2: I suggest to re-define all acronyms within each sub-section here - but this is just a suggestion and may also be left to the editorial staff

- line 128: CONUS domain?

- line 133: 12Z and 03Z?

- section 2.4 considering that the observations by CrIS are affected by the smoke, and that this is not accounted for in the CrIS retrieval algorithm (at least I would guess it is not), how confident may we be in the retrieved surface and atmospheric temperature and atmospheric humidity for the smoky observations? 

- section 3.1: I would appreciate here some introduction about the expected effect of the studied phenomena (coarse / giant particles, pyrocumulus) on the radiance in the different channels used in this work; that would allow the reader to follow the developments and understand the conclusions more easily

- figure 2: what are the Z after the hours? (as line 133)? Is this local time, or UTC, or other?

- section 3.2: Again here I would appreciate an introduction about the signatures of gases in the TIR atmospheric window, especially the fact that they are rather small; maybe using cross-sections also. Within the GOES 10.35µm and MODIS 11µm channels, one would indeed find some weak WV absorption bands, some very weak CO2 absorption, relatively intense O3 absorption, possibly NH3 absorption if some is present but no N2O or CH4 absorption (no band in those channels). Because in that spectral range gas absorption is rather low, no increase within physical range of any of those gases would lead to 25K BT reduction at 11µm. This is also the conclusion that the authors reach after experimentally testing that hypothesis, and the proposed approach is also important and interesting but I think that the physical base should be discussed as a complement.

- lines 295-298: could you elaborate on what might be the reason for this difference? Would this be a sign that the CrIS "smoky" profiles are not "close to reality" because the smoke impacts the observation and this is not accounted for in the retrieval?

- figure 4 (j): where is the orange line? I would guess exactly behind the green but I would mention it in the legend, or redo the plot with e.g. different line widths so that the line can be seen.

- section 3.3: again a bit of physical basis here would be useful, I think. The expected BT in channels around 11µm, in absence of clouds / aerosols (and at relatively low viewing angles), would be slightly lower than the surface skin temperature (unless looking above a surface with low emissivity, of course). This is indeed what the radiative transfer shows when using either the CrIS skin T from "clear" or "smoky" cases (but was this CrIS skin T "right" for the smoky pixel?). One question that remains non-addressed in this section would be if it is reasonable that the surface (skin T) would cool by 25K in a short time if under a thick smoke plume. I guess this could be done by looking at other days (without smoke) BT daily cycles and how much night BT differs from day BT under relatively similar circumstances (except the smoke), and how long the surface takes for cooling after sunset, for example.

- figure 5 (b): time is here given in local time while I think almost everywhere it was UTC time. This should be consistent and my preference would be to have both UTC and local time on each plot (local is interesting to know which part of the day we are looking at, while UTC time is interesting if the reader wants to compare with any other data)

- lines 382-383: again some physical explanation here would be nice - indeed the atmospheric temperature has only a second order impact on the observed BT, being through the atmospheric gases thermal emissions, which depend on their temperature and cross-section - the latter being rather low in the used channels

- line 410 and following: I am not so sure one can say that there is no noticeable cooling in the plume area - the plume is widespread and there's some widespread "reddish" area in VIIRS 10.76µm that seems to match the grey area in VIIRS DNB. However this "feeling" might come mostly from the fact that a completely different BT scale is used for that plot with respect to the 2 "day" plots. I would strongly suggest using the same scale, or at least the same range of temperatures for the color scale (currently 70K for daytime and 20K for nighttime)

- lines 434-436: is this also true at night?

- lines 439-441: those numbers are rather different from the numbers given lines 430-436: +80 -50 W/m2 do not compensate, while -2 and +1.9 W/m2K almost exactly compensate. Am I missing something? Or is this within error margin?

- Figure 7a (and c to a smaller extent): I find it rather hard to really see the linear relationship in these clouds of points. In Figure B it is much more clear. Are you sure that a linear relationship is expected between the SW flux and the TOA LW 11µm BT? 

- lines 455-457: this sentence is a bit too straightforward and maybe misleading. The characteristics of the plume can not be retrieved based on the TIR channels, at least this is not what the manuscript is about. Maybe the authors could say that plumes could be identified from the observed BT changes in the TIR, after additional work allows discriminating the reason for those BT changes (clouds, Ts changes, smoke).

- conclusions to be updated depending on changes in the manuscript

This is a very interesting study focused on the longwave (LW) signatures of a thick smoke plume from the 2021 Dixie Fire in California. The satellite remote sensing aspects, radiative transfer modeling, and surface station analysis are generally well constructed. The narrative is well-written and organized. My only major comment relates the coarse and giant particles analysis in Section 3.1, which is critical to the conclusions of the paper.

I agree with the comments by other reviewers on incorporating weather radar data to provide a better representation of the horizontal and vertical extents of large smoke debris (ash and other pyrometers). The Dixie Fire plume should have good coverage from radar sites in the region. Adding these data into the current analysis will provide a better constraint on variables driving the observed cooling.

Reflectivity and correlation coefficient (CC) provide a quick and definitive way to examine the presence (or lack thereof) of both hydrometeors (high CC) and pyrometeors (low CC) in the plumes. Smoke plumes examined with radar in previous studies coincided with radar echoes 20+ km downwind, indicative of large particles far from the fire (e.g., McCarthy et al, Lareau et al, Peterson et al.; see example papers below). A quick check of the meteorology on the dates examined here reveal relatively strong winds in the mid-troposphere, which would likely facilitate transport of these larger particles, perhaps as far as the ground station sites. In addition, the analysis period appears to coincide with very intense fire behavior, which would result in higher altitude injections of larger pyrometeors.

https://doi.org/10.1029/2019GL084305

https://journals.ametsoc.org/view/journals/bams/103/5/BAMS-D-21-0199.1.xml

https://journals.ametsoc.org/view/journals/bams/103/9/BAMS-D-21-0049.1.xml