This study analyzed two datasets from collaborative measurements with the EISCAT UHF and VHF radars in Tromso, Norway, during moderately active conditions characterized by hard particle precipitation. By applying the difference spectrum fitting method developed in a previous study, the height profiles of the ion-neutral collision frequency in the MLT region were derived. The ion-neutral collision frequency is a critical physical parameter for understanding the ionosphere-thermosphere coupling process; however, its features have not been fully elucidated. Based on the derivation, this study presented the impacts of precipitating particle forcing on the ion-neutral collision frequency and discussed plausible causalities that produce variations in the collision frequency. However, the explanation of the calculation results and physical mechanism requires more rigorous and comprehensive analysis of data and consideration from multiple perspectives, as the current text is biased toward only certain aspects or interprets phenomena in an overly simplistic manner. These issues are summarized in the major comments. The accuracy of the calculations and discussion is significantly low, and the results of this study do not meet the standards for publication. Since the quality is unlikely to be improved by reanalysis, as mentioned in Major comment 1, the recommendation for the editor is to reject.

[Major comments]

1. Artificial discontinuity at 100 km altitude

In examining Figure 2b, a discontinuity in the collision frequency is observed at approximately 100 km altitude. This discrepancy is more evident in the line plot presented in Figure 3. The quick-look figure of the EISCAT measurements available in the Madrigal database clearly indicates a significant difference in the noise level of the UHF-measured electron density above and below 100 km. One of the conclusions of this study addresses the differential effects of thermospheric density variations at altitudes above 100 km. Based on the results presented in Figures 4b and 6b, this study corroborates that thermospheric density decreases and increases at altitudes below and above 100 km, respectively. However, the validity of this conclusion is subject to scrutiny given the apparent disparity in the quality of the EISCAT measurements, which constitute the primary data used to calculate the collision frequency.

While the manda pulse code employed in radar measurements exhibits an advantage in measuring lower-altitude electron density with high range resolution, it demonstrates reduced accuracy in measuring ionospheric parameters in the E region and above compared to other pulse codes. The analysis presented in this study does not account for this characteristic (at least, no explanation is provided in the text).

As illustrated in Figure 3, the collision frequency below 95-100 km altitude exhibits a strong dependence on a priori. The authors assert that the altitude region where the influence of the a priori is significant extends up to 95 km, and they have hatched this region with gray shadows (Figures 4 and 6). Figure 3 demonstrates that the upper altitude limit is contingent upon the choice of a priori, and the a priori should be considered dominant up to 97-100 km altitude. The noise level of the EISCAT measurements is elevated for collision frequencies above 100 km altitude, as noted above. Given these considerations, the calculated results for any altitude range are deemed unreliable for this dataset, and there appears to be no justification for the discussion and conclusions based on these ambiguous results. Consequently, the results, arguments, and conclusions drawn in this study based on the observed data are considered unreliable. It is recommended to utilize EISCAT UHF and VHF simultaneous observation data employing another pulse code rather than manda.

2. Physical parameters to affect on the ion-neutral collision frequency

The collision frequency is proportional to the thermospheric and ionospheric-plasma densities, as demonstrated in Equation 1. However, this equation represents a simplified model and is also dependent on temperature (Prolss, Physics of the Earth's Space Environment, 2004). Given that the event was observed during periods of high geomagnetic activity, it is logical that the temperature of the thermosphere and ionosphere increased due to particle heating and Joule heating. Rather than restricting the analysis to attributing all collision-frequency variations to the density variations, it would have been more comprehensive to incorporate temperature effects in this study.

The variation in thermospheric density is estimated from the increase or decrease in collision frequency (Figure 4b). As demonstrated in Equation 1, the collision frequency is a function of thermospheric density; however, it is also dependent on ionospheric plasma density. Lines 167-170 in the text briefly describe the derivation method, yet it remains unclear whether the thermospheric density was calculated considering the electron density measured by the EISCAT radar. If this factor had not been considered, it should have been incorporated into the calculations. If the collision frequency has been calculated taking this into account, the error in the atmospheric density should be determined by considering the error in the measured electron density and discussing its significance in relation to the magnitude of the thermospheric density variations.

3. Feature of the vertical motion in the lower thermosphere

It was previously mentioned that variations in collision frequency were not exclusively attributable to fluctuations in the thermospheric density. Even if density variation is presumed to be the primary factor governing the collision frequency variation, the explanation provided in the text would not align with the characteristics of vertical motion in the lower thermosphere. When examining the vertical motion of the lower thermosphere, it is essential to consider horizontal motion, particularly along isobars. In the lower thermosphere, where horizontal motion predominates, an upward displacement of the isobar in the heated region induces an apparent vertical component of the wind in geographic coordinates due to thermospheric winds flowing along the isobar. In instances of intense localized heating over brief periods, upwelling across the isobar may occur, but this phenomenon dissipates rapidly in conjunction with vertical oscillations. Upon cessation of heating, the isobar expansion terminates, and the apparent vertical component diminishes. However, during a transition process under force balance between buoyancy and gravity, the atmosphere undergoes oscillation (i.e., atmospheric gravity waves are generated), and vertical motion may persist for a duration. Nevertheless, this phenomenon does not result in an increase in the spatiotemporal mean density.

Figures 4b and 6b indicate that the atmospheric density increased above 100 km altitude irrespective of the electron density level at 95 km altitude employed in this study. During periods of high geomagnetic activity, characterized by high electron density, the atmospheric density above 100 km altitude may increase. However, the intermittent increase in electron density over the two and a half days of December 13-15 suggests that it is improbable that the energy flow from the magnetosphere to the polar thermosphere/ionosphere is sustained at a sufficient level to support the density increase. Considering that particle heating should have occurred at the same location as the aurora, and given the likely structured nature of the aurora, it is implausible to assume constant upwelling in the EISCAT radar beam, although Major comments 1 and 2 elucidate the unreliability of the calculation results. Even if the calculation results capture some degree of nature, the physical interpretation of this study cannot be objectively substantiated.

[Minor comments]

L3: "momentum transport" should be revised to "momentum transfer."

L13: "different a priori collision frequency profiles" may be better to say "various a priori collision frequency profiles."

L48-49: A magnetometer in Tromso, which can be checked at the IMAGE webpage, presents substorms during the high electron density periods selected in this study. The effects of Joule heating should be discussed, along with those of particle heating.

L60-61: According to the EISCAT QL from the Madrigal database, the transmitter powers of the UHF and VHF radars have not reached these numbers.

L66: "06-15 UT is analyzed" should be revised as "06-15 UT, is analyzed."

Section 5.1: The effects of the ambiguity of the beata parameter on the derived collision frequency should be evaluated in a quantitative manner.

L222-224: The explanation is incorrect, according to Figure 3.

Section 5.5: The meteor radar measures the horizontal wind, but what this study mentions is the density. To apply the meteor radar measurements, experimental evidence or theoretical support is required to link the wind and density in advance.

The authors use simultaneous measurements from the EISCAT UHF and VHF incoherent scatter radars to examine the dependence of the ion-neutral collision frequency on particle precipitation. They purport to show a significant impact and that the particle heating causes a significant uplift of neutral gas between 90 and 110km.

This was a fascinating and worthy attempt to use the multifrequency data to extract information about the underlying neutral atmosphere.  The technique and type of analysis involved was sensible and interesting.  However, inspection of the underlying data has made me concerned as there seems to be some systematic noise issues that may persist even at high density levels and need to be properly identified and discussed. I feel the current discussion is not enough and misses some worrying features in the data (e.g. the UHF noise band from 100-110 km, the heightened noise spikes at all altitudes in the VHF).

I think that the fairest approach would be to ask the authors to conduct an analysis of the underlying data noise, speak to EISCAT staff about the measurements to check if there were additional issues (the system temperature on the VHF seemed rather high at times). This will give everyone more confidence in the data and the analysis. This could be done through rejection and request resubmission, though I am minded to recommend a major revision instead, unless the authors feel the timescale to carry this out would be too long?  Even if there are fundamental issues with the underlying data there may be solutions for mitigating these that could make the study worth publishing even as those flaws are acknowledged.

Major comments

Data noise:

The underlying data as presented in figure 2 is somewhat noisy.  I worry this may impact severely the analysis that the authors have done.  Did they consider increasing the integration time and/or size of the range gates?  I note that the noise issue is most pronounced in the UHF data, inspecting the MADRIGAL summary plots for 13 -15 December 2022) the UHF shows a persistent band of increased noise from 100-110 km altitude, which is worrying.  The VHF shows highly noisy data below 110 km.

One would expect an increased level of relative noise in regions of lower density (i.e. low altitude) but for the data here it seems remarkably high with large multi-gate spikes. It is harder to tell if this persists when density is high though I note in figure 2b the first large precipitation event after noon on the 13/12/2022 there is a noticeable step change in the collision frequency that continues through the period. It appears to reduce during periods of very increased density, though this could be a color scale effect and so I am still wary.  That the discontinuity manifests in the collision frequency data from the difference spectrum fitting method as well as in the GUISDAP analysed data suggests that it is an underlying data issue and not simply an analysis problem. If this is the case it would impact all of the data and so restricting analysis just to the high-density regions would not completely mitigate the problem. This needs thorough investigation.

I wondered if this may be that this is simply indicative of the transition from the a priori value to the measured value. However, this does not resolve the issue of the band of noise in the UHF density data that is suggestive of a bigger problem (and in the VHF manifesting in a different way).

The issue with the data is harder to see in May 2024, though the low altitude VHF data still shows unusual noise structure. I had a quick look at some older manda data in the Madrigal archive and the noise structure looks different (this is anecdotal and far from comprehensive), which makes me suspicious of these recent observations and whether there is an underlying issue in the latest version of the manda code. Did the authors discuss the data with the EISCAT staff? It would be worth identifying if this is a problem with the current implementation of manda or something else (some sort of aliasing problem with more contaminated range gates at higher altitudes?)

I think it is essential for the authors to conduct an analysis of the uncertainties and the level of variability (noise) in the data to identify any trends, or sharp changes with altitude and underlying density level.  Expanding the analysis upwards beyond 110 km might be instructive in terms of identifying any additional discontinuities that would point to a problem in the noise level.

Minor to moderate comments:

Line 24: the assumption is made that the ion density is equal to the electron density; this is generally true at higher altitude but as one moves lower in the D region the negative ion density increases such that it may become appreciable (especially in darkness). How would this affect the use of Equation 1.

Line 69: “the UHF and VHF radars were operated in the manda zenith mode, also known as the Common Programme (CP) 6” – this is not accurate, it is similar to the CP6 mode.  “Common Programme 6” has a very specific meaning beyond the pulse code and pointing direction. It would be more correct to say that “the UHF and VHF radars were operated pointed in the zenith using the manda pulse code, which is optimised for high resolution D-region measurements (as used in the EISCAT Common Programme 6)”

Line 130: a little more information on how the EISCAT climatology is constructed would be welcome to the reader who may not be familiar with EISCAT and climatologies. Given the way in which EISCAT data presents more explanation on how it is a straight line would be welcome.  I assume this also indicates that the standard EISCAT analysis is not using NRLMSIS to provide the a priori data to the fit otherwise the two lines should be the same. Apologies if I have got this muddled but I think that demonstrates the need for further information on how this is constructed.

Line 164: The authors suggest that the detected uplift is due to direct particle heating. This may play a sizeable role; however, did the authors consider the role of Joule heating? This tends to maximise above the region of interest in this work but Baloukidis et al., 2023 (https://doi.org/10.1029/2023JA031526) showed that there can be appreciable heating towards the top of the region of interest and  Deng et al., 2011 (https://doi.org/10.1029/2010JA016019) suggested that the uplift can start at lower altitudes. I think it unlikely that Joule heating alone could cause the potentially observed upwelling at the reported altitudes, but it is worth discussing this possibility with reference to the potential mechanisms and qualified assessment.

Line 189: “significantly” have you tested the result for significance. For example, a Kolmogorov-Smirnov test comparing the underlying data behind each profile with each other.

Line 199: “the difference between both profiles is within the measurement uncertainty and conclusions have to be drawn carefully” is a very sensible statement but have the authors attempted any significance testing of their results.

Line 205: I wonder is there an independent neutral temperature measurement over EISCAT at this time?  There have been temperature mappers and lidar focused on the region though I admit I am uncertain of the data range that exists.  Alternatively, is there any satellite data for the interval that could be used?

Line 222; the jump (or rather a kink in the profile) at 110 km appears in both cases in Figure 3, it is reduced when NRLMSIS is used but it is absolutely still there. Therefore, it is erroneous to state that “This jump is only found when the standard EISCAT… not when NRLMSIS climatology profile is used”.  Please, correct this. I would also note that the jump is in all time sectors (fig 3a) but more pronounced in the noon and dusk sectors. The fact it persists at all times may suggest that the SNR problem is not just constrained to low density periods. In fig3b the jump is most pronounced in the dusk, then the noon and is indeed much reduced (to vanishing) in the two other sectors. This needs some thought and explanation; I suspect it may be connected to the SNR issue and how using NRLMSIS will mitigate that to a degree at higher densities, but I am unsure.

Line 245: fluxes and ionization rates are often presented in units of per cm cubed, but I recommend SI units. Leaves less room for mistakes. Same for neutral densities in Figure 6.

Lines 266 and 270; essentially saying the same thing twice in quick succession.

Lines 264 onwards: I am not clear on how the analysis of the tidal structures leads to the conclusion on line 275. Please provide more detail of the underlying mechanism.

Figure 6b: why not include error bars on the red line?