Review on “Deriving column-integrated thermospheric temperature with the N2 Lyman–Birge–Hopfield (2,0) band” by Cantrall and Matsuo.

This paper presents a new method for retrieval of lower thermospheric temperatures using N2-LBH (2,0) band. It compares the retrieved products with GOLD level 2 temperatures, simulated data, and MSIS data. Overall, I find this manuscript interesting and can a potential AMT paper after addressing some major concerns listed below.

Major Comments:

1. Lookup Table (Lines 161-162): The lookup-table data has not been provided anywhere. I would suggest including them as a part of publicly available data, so that someone interested in reproducing/verification by independent means can use/validate them.
2. PCA of simulated LBH emissions: Line 115: “The second leading….(.. explained later)”. This is not clear to me how the 2^nd leading mode would only contain the temperature variability. Why would it not contain, for example, the geomagnetic. variability? Can you use a bunch of simulated spectra corresponding to temperatures in the 300-1500 Kelvin range and show that the 2^nd leading mode is associated only with temperature changes?
3. Line 126: Shot noise: It is said that the spectra are just simulated/model/synthetic spectra. How can a model/simulated spectra will contain shot noise? Are you using a set of spectra or introducing some random noise in the spectra and then calculating the shot noise? Please add more explanations.
4. Lines 243-246: Variation in wavelength registration: Better used an atomic line but try to avoid OI-135.6 nm as it is very strong emission and on occasions degrades the detector. Variation in wavelength resolution: Again, better try to use some atomic line other than OI-135.6 nm.
5. Line 183-185: GOLD case study: Why you are not using the errors available in the L1C data? Why do you need to simulate the error?
6. Line 224: “T_{ci}^{G} is also….based on the SZA.” In the previous section it is stated that sampling at peak altitudes introduces 30-90K error. Then why are you using MSIS sampled at peak altitudes. I would recommend calculating GOLD equivalent effective temperatures using MSIS profiles and contribution functions from radiative transfer model. It will give better comparison with GOLD L2-Tdisk, particularly with version 3 TDISK. This can be presented as an additional row in the comparison (Figure 5).
7. Section 4: The authors used the unbinned data from an old release version-2 (V2), which I cannot locate in the two GOLD repositories provided in the data availability section. As there is poor signal to noise (SNR) concern and potential bias concern, I would suggest revising the analysis and results with version 3 (V03) GOLD TDISK and 2x2 binned L1C data (L1C-V03). Specifically, revise Figure 5 with the V03 data.
8. Lines 277: Previously the authors mentioned that this retrieval is unaffected by biases in emission intensities as the absolute values are not important, but the spectral shapes are. Then why would systematic errors in intensities, which will basically introduce some bias, would introduce bias in temperature calculation? This also contradict the conclusions in lines 318-322, which says absolute band intensities are not required.

Minor comments:

Line 107: The sentence may be revised for clarity.

Line 180-181: Provide reference?

Line 242: Full disk measurements goes on until 23 UTC.

Line 288: What is the x-axis in figure 7? Is it local time at all longitudes or local times at fixed longitude?

General comments

• The technique for determining temperatures from disk LBH emissions that is described in this manuscript represents an important new tool that may enable both further analysis of data from GOLD, but also potentially future instrumentation. The method has several strengths over existing techniques, including not requiring knowledge of the absolute brightness of the emission, not requiring a broad portion of the LBH bands to be sampled, and not requiring the kind of spectral resolution that has underpinned some other techniques. As such, this work should be of great interest to those interested in thermospheric observations, and techniques for analyzing such observations.
• The manuscript includes a good description of the uncertainties, related to instrument wavelength and noise.
• The detailed description in Section 3 of how the disk temperature should be interpreted as column temperatures is particularly important and as these and similar data utilized by a broader community this kind of consideration is essential.
• The particular case study, utilizing data from multiple spacecraft and centered around a moderate geogmagnetic storm provides a good demonstration of how the temperatures retrieved from the technique introduced here vary under such conditions, and demonstrates their utility to the broader scientific community.

Specific comments

1. Line 90 – Is the factor of 1.6 mentioned here an issue with the current approach? My understanding form later sections is that it is not. If this is the case, I believe it would be worth explicitly stating that here.
2. Line 177 – Is this statement true, if the model for LBH with temperature is imperfect?
3. Line 180 – I believe that the O₂ absorption cross-section also varies (albeit not strongly) as a function of temperature. This will further complicate this factor, although it is likely still minor.
4. Line 182 – It is certainly true that the shot noise, which is proportional to the square root of the emission signal, is a major part of the instrumental noise. However, particle noise is, at least at some times, an additional random noise source. Importantly, it’s behavior is not the same as the shot noise as it is unrelated to the brightness of the signal being observed. See for example the description of the particle background and its associated flag in GOLD Release Notes Revision 4 - https://gold.cs.ucf.edu/wp-content/documentation/GOLD_Release_Notes_Rev4.1.pdf. This may, potentially, be an important consideration in the case study presented in this manuscript.
5. Line 267 – The east-west gradient that is described here is not clear to me in Figure 5. I would recommend that this be demonstrated more clearly, perhaps in a line-figure such as Figure 6, as I believe it is an important point that current, at least I struggle to see from the image.
6. Figure 5 – The range over the disk where T_ci_G appear is smaller than that of Tdisk. Is the origin of this a differences in the solar zenith angle ranges, or some other criteria used in the approach described here that differs from the publicly available Tdisk?

Technical errors

• None noted.