The technical note written by Ellis Remsberg makes a valuable and insightful contribution to the current research on the variability and trends of stratospheric water vapor, particularly in relation to the comparability of satellite-based (HALOE) and in-situ-based frost point hygrometer (FPH) observations above Boulder. The FPH method is assumed to be the most accurate. The paper presents several new and interesting findings, including:

(i) Identification of problems with the HALOE sampling at 40N from 2001 to 2005, particularly during late winter and spring.

(ii) Demonstration of good agreement between the trends derived from HALOE at 30 hPa and FPH data, by considering a limited time span from 1993 to 2002.

(iii) Giving few hints that both FPH observations and HALOE closest data points can be strongly influenced by moist polar air isentropically transported as remnants of the NH polar vortex during late winter and spring.

Overall, the technical note is an important addition to the current research on stratospheric water vapor variability and trends, and provides new insights on the comparability of satellite-based and in-situ-based observations. Here, few detailed comments:

Main points:

(*) To improve readability, it may be helpful to present Figures 1-3 on the left panel and Figures 4-6 on the right panel of a single figure, allowing for easier comparison between the HALOE and FPH data at each step of the data handling process.

(*) The correction procedure discussed by Scherer et al. (2008) for FPH data is not applied in this study. As the trends for HALOE and NOAA appear different in their Figure 6b compared to the current study, it should be noted that the current study utilizes slightly modified HALOE data by excluding data from 2001 and 2005 and FPH data without the correction discussed by Scherer et al. (2008). Further discussion on the potential impact of this correction on the results would be valuable.

(*) While the comparison at 30 hPa is important, it would be informative to also compare the trends at 70-100 hPa, which have been widely discussed for trend analysis in the past (Solomon 2010, Hegglin 2014, Konopka 2022). These studies have shown strong positive trends from FPH observations for the period 1993-2001, whereas other data used in these studies (UARS HALOE shifted to MLS, merged satellite, SWOOSH) have exhibited weaker or even negative trends. Examining these pressure levels may provide additional insight into the ongoing discussion of the trends.

(*) My most concern is related to the comparison between your Figs 3 and 6, i.e. between HALOE and FPH based trends for the period 1993-2002 at 30 hPa. Yes, even if the trends are very similar, the variability derived from the MLR (thin solid lines) is very different (i.e. thin line in Fig.3 compared with the thin line in Fig. 6). Why?. I guess the correlation between these two thin solid lines is very weak. So why we should trust the trends derived from these two so different lines?

(*) In the second part of your paper, you discuss isentropic polar intrusions into the mid latitudes as potential reasons for additional variability both in FPH data and collocated HALOE observations. Following your analysis, this effect should be expected mainly in the late spring and winter. This is certainly an interesting point not well discussed in the past. However, as discussed in Konopka et al., 2022 (see their Fig 4d and 4e), the winter/spring effect on the trends after 2000 was mainly diagnosed in the tropics and not in the Boulder region. Thus, "intrusion effects" on HALOE/FPH observations around 40N and the largest trends after 2000 during winter/spring seasons are probably two different issues.

Minor points:

L65:

"auto-regressive effects" - too technical here, please explain it

L70:

please explain the abbreviations "SS", "SR"

L112:

"have a significant (CI)" ?

L200:

"....from the LIMS experiment" please explain the abbreviation

Review of Remsberg et al:

Given that there is only one long-term balloonborne stratospheric water vapor record from Boulder and few satellite measurements from the 1980’s and 1990’s, understanding and interpreting the trends present in these records is extremely important. One of the more puzzling findings that has emerged from the literature is whether or not the positive water vapor trend observed in the Boulder water vapor record is real. Several studies have highlighted that satellite measurements over the late 1990’s/2000’s (i.e., HALOE and SAGE II) don’t seem to show the positive trend present in Boulder data. It has been suggested that the trend difference is due to some unknown drift of the Boulder record, or perhaps that the measurements from a single location are not representative of the true zonal mean. Within this context, the present work seeks a better understanding of these differences. The works shows that prior to 2002 the HALOE and Boulder records are in reasonable agreement, and provides some evidence that the change in HALOE sampling in late winter/spring after 2002 could contribute to the divergence of the trends between HALOE and Boulder when considering the longer time period through 2005. I find this idea and the presented analysis intriguing, but the work seems somewhat incomplete. I believe that more work could pretty easily be done to test the hypothesis that the trend difference is due to sampling changes in HALOE. Below, I provide some specific comments on the manuscript, and offer a few suggestions geared towards improving this analysis to make it better able to address the hypothesis it proposes and the underlying questions regarding trend differences between HALOE and FPH.

General comments:

• If the continued positive trend in WV at Boulder for and end time of 2005 (as opposed to 2002) is caused by vortex air sampling, shouldn’t this show up in equivalent latitude changes in Boulder? This is indeed what Scherer et al. 2008 attempted to include, and they were unable to see any change that could explain the WV trends at Boulder. You could probably look at equivalent latitude as a predictor, or just changes in equivalent latitude at Boulder, to address this. On a related note, Lossow looked pretty thoroughly at the representativeness of Boulder and concluded that there was a real discrepant trend between HALOE and FPH that couldn’t be explained simply by the fact that the FPH measurements come from a specific site. At the very least, some more thorough discussion/comparison to the work by Scherer and Lossow is warranted here.
• If the HALOE and FPH trend differences are explainable by sampling of vortex air at Boulder during late winter/spring, then one could pretty easily look at seasonal trend estimates to see if the agreement is better during other seasons. This can be done very easily with MLR (see, e.g., the LOTUS activity for ozone).

Specific comments:

Lines 44-45: This statement about the Lossow paper seems incomplete and should be more nuance, in my opinion. They showed that Boulder is representative of the zonal mean for trends over a few decades (e.g., late 1980’s to 2010), but that at a decadal scale the trend differences between a single site and the zonal mean could be different.

Line 63: It would be nice to provide a brief summary of your MLR  hereso that readers don’t have to dig into multiple other papers, and to be as transparent as possible about what you’ve done. I see that the MLR is described in the next paragraph, so maybe you could just mention that it is described in detail in the next paragraph.

Line 69-70: You say HALOE WV values for SR events are larger than SS for 2002-2005, but the following sentence doesn’t actually explain why. Please explain.

Line 80-82: You are using a 70° wide longitude range around Boulder and say that it can resolve wave-1/2 features in the SWV field. The 70° range represents 1/5^th of the longitude values so I don’t see how this is possible.

Line 82: Is the MLR model run with monthly means or daily data?

Line 83: What do you mean the QBO is approximated as a 28 month cycle? Are you not just using a QBO index like the Singapore winds at a certain level (or two), or the NCEP indices? Please explain.

Line 86: I assume the AO is also significant? It is not listed here.

Line 89: I appreciate that the trend units of %/decade are useful, but can you please also give trend numbers in terms of ppmv/decade here and throughout the paper. A number of other studies (most, I thnk) have expressed water vapor trends this way, so it makes it much easier to cross-reference and compare with other studies.

Line 95-96: What do you mean about the MLR terms being non-orthogonal due to uneven spacing of the data? This doesn’t make sense to me.

Line 98-99: I really think the first description of this drop was from a Randel et al 2006 paper, so that should be cited here.

Line 110: You interpolate FPH profiles, but do you apply any smoothing? I think it is appropriate to do so here, particularly if you are going to make any comparison between the noise/scatter of the FPH and that of HALOE as in Line 117. As FPH is much higher vertical resolution, it needs to be degraded to something approaching the HALOE resolution in order to make a reasonable comparison.

Line 117: See above comment.

Line 119: The SPARC 2000 report is pretty out of date at this point, and improved FPH estimates have been provided by Hall et al. Please cite this and use numbers from that paper, as they superceed anything that might have been in the old SPARC assessment.

Hall, Emrys G, Allen F Jordan, Dale F Hurst, Samuel J Oltmans, Holger Vömel, Benjamin Kühnreich, and Volker Ebert. “Advancements, Measurement Uncertainties, and Recent Comparisons of the NOAA Frost Point Hygrometer.” Atmospheric Measurement Techniques 9, no. 9 (September 5, 2016): 4295–4310. https://doi.org/10.5194/amt-9-4295-2016.

Line 145 (and elsewhere): These uncertainties seem extremely small to me. 1% is something like 0.05 ppmv, and the residuals (e.g. in Figure 3) are ~0.5 ppmv, so it seems unlikely that the trend values are really this precise. Has lag-1 autocorrelation of the residuals been taken into account for computing the trend significance? If the error bars are significantly larger, it would substantially change how one views the agreement (or lack thereof) between HALOE and FPH.

Line 162: What does “qualitatively accurate” mean here?

Figures:

A number of figures are really zoomed out on the y-axis, making it hard to see the data. E.g., in Figure 1 the data range goes from ~4.5-5 ppm but the y axis goes from 3 to 6 ppmv. Please zoom in the axes here and elsewhere.

A number of the figures would make more sense if they were combined. E.g., Figures 1-2 together, 4-5, 9-10.