General comments

This paper investigates the fingerprints of Covid-19 on 41 elevated mountain sites over the world, mainly in the USA and Europe. The scientific interest of the paper is very important, regarding the ozone chemistry related to sources and sinks. The paper is excellently written and well organised with the different chapters.

The measurements sites and the methods for data selection are well described and the data selection is accurate, with night-time values or daily 8h maximum averages for some stations. The use of IASI data is a good choice for comparing with satellite data.

The quantification of the 2020-2021 anomalies is well explained and discussed, related to the emissions reductions shown in Table3.

The conclusion is robust, due to the high number of sites and the O3 reduction is comparable to the IASI data.

All figures are excellent quality, easily understandable and well commented in the text. The supplementary material is also excellent quality.

This paper is suitable for publication.

Minor comments

The author should very briefly discuss about a possible reduction of stratosphere/troposphere exchanges in the period, as a non-negligible part of free troposphere ozone is coming from the stratosphere and as the stations are located in altitude.

This manuscript calculates trends and anomalies in ozone concentrations at high elevations stations and explores the response in these metrics due to changes in behavior during the COVID-19 pandemic. This is additionally explored by examining temporal profiles in satellite retrievals of O3 column data.  

Overall, the paper is a thorough and well written account of the changes experienced at the chosen sites. Subject to some clarification of the methods used and the specific comments below, this paper should be accepted for publication.

General Comments

In sections 2.1.2 the authors describe the calculation of the O3 anomalies after detrending and de-seasoning the timeseries. I have two primary concerns with this section:

• While I believe the process described here is sound, my concern is that as written, reproducing the methods from this description is not facile.
• There is substantial mixing of mean averaging with median seeking methods (i.e. quantile regression at 50%).

Restructuring of this section would go a long way to allay these concerns, and I would consider the inclusion of a simple flowchart in the SI making it overtly clear which steps are applied in what order. A good example of the language that is hard to follow surrounds L114: “Last, we used the differences calculated in the previous step…”. Are these differences referring to the resulting de-seasonalised timeseries, which makes sense for a trend, or is this referring to the climatological year, from which one could feasibly calculate anomalies – I believe the authors are referring the former, but I hope this illustrates the uncertainty that is introduced throughout this section. Being explicit with the use of “mean” or “median” over “average” would also help the reader keep the steps clear.

On the second point, my major concern is that the climatological year has been calculated via mean averaging, but then timeseries derived using this to remove seasonality have their trends defined by the median (via QR). After Chang et al, 2023 trend analysis using QR is the preferred method here, but I would question why de-seaonalisation was not conducted using a climatological year calculated using the median also, as one would surely want this to be less sensitive to outliers in any given year.

Specific comments

Line 101 – “The deseasonalization allows to produce a more…” Should perhaps read: “The deseasonalization allows the production of a more…”

Figure 2. – The use of two colours per p-value is not well described. Is hue used to denote trend sign, and saturation for significance? In the previous figure a very similar colour pallet was used to show region, which has now been moved the shape in this figure. If this is the case, the use of a different colour pallet here for significance (one colour only) and allowing the x-axis to denote trend direction would be much clearer.

Figure 3. – Referring the reader to fig. 1 for the definitions of the colours is not good, as the figure + caption should stand alone much more readily. The regions could be added to the y-axis to make the groupings clear.

Line 194, relating to figure 3. The wording here could be changed as I don’t think “clearly shows widespread… …in 2020” is strictly true. A similar statement could be made about 2009 or 2015.

Line 217, 2021 falling in the top 5 of 18 years – essentially means 2021 falls in ~ top 1/3^rd of years over that period? This could be more clearly phrased.

Table 3. This table is attempting to show too much information at once, could be separated out into two to avoid the use of the parenthesis notation, which interrupts being able to read down the columns clearly.