This manuscript studies geomagnetic K-indices measured in several observatories in Norway. The main author has digitized earlier indices from observatory yearbooks and IAGA Bulletins for tens of years, and calculated indices for the recent years using digital data. A detailed comparison is made for the early K-indices produced using hand-scaling and by two digital methods, the FMI K-index method and the local TGO K-index method (and its "preliminary" version). The manuscript studies the K-indices for several Norwegian observatories and discusses their long-term homogeneity. Authors have completed quite an extensive effort and, overall, the manuscript contains interesting information on the different K-index versions and methods to produce them. Therefore, the manuscript, eventually, may deserve to be published. However, I have several questions and even different interpretations on the results. Therefore, I have to suggest considerable modifications to the contents of the manuscript.

Presentation of the production of the digital K-indices using the two (actually, three) digital methods should form the solid basis of this work. Unfortunately, this basis is not laid very well and must be considerably improved. Authors say they used the Fortran code from IAGA to calculate the FMI K-indices. However, later, it becomes clear that this is an outdated code version which contains errors. Authors also elaborate later in the manuscript some problems due to those errors. The code has been corrected by FMI already a long time ago, implemented in a newer c code. It is not clear to me why the authors have used the outdated code version. If it is due to unfamiliarity with c language, this is no problem since codes can easily be transformed from one language to another by AI. I am confident that authors should indeed use the correct(ed) code. This will set all results on a more solid basis.

Readers must be able to understand the different K-index methods in order to be able to evaluate the differences between the different K-index series. Now, the description of the FMI K-method is too short and unclear. E.g., it is not understandable how "These preliminary K indices are then used in calculation of the first fitted QDC" and how "These K indices are used in calculation of the final fitted QDC". This, and the whole recipe should be clarified, perhaps, using examples or figures in order to make the manuscript complete and understandable. It should be clear to the reader what sets the absolute level of the QDC curve and how its form will get determined. Also, I would like to have the problems related to the FMI QDC discussed here, not much later in the manuscript.

Even the description of the TGO  method on pages 7-8 needs clarification. I dislike another concept of QDV, meaning the "values" used to calculate the TGO K-index, while they obviously are just hourly values of the (monthly) QDC. If this interpretation is wrong, please clarify. Else, simplify notation and remove QDV.

 I read that the H-component is used to calculate the QDC, so obviously TGO method uses this component for K-index calculation, although I did not find this clearly mentioned. On the other hand, page 7 says that the FMI method uses both X and Y-components to calculate respective K-indices (selecting the larger of them). The question is then, do authors compare K-indices that are calculated from different components in the two methods? As authors describe in the history part, during the hand-scaling time, both components were used (even Z originally), so the FMI method is more in line with long-term homogeneity. So why does the TGO method use the H-component, as it should be equally easy to continue the historical procedure? These issues are not much discussed in the manuscript.

I would also like to see a more complete description of the TGO' method. (A better name would be, e.g. TGOp, p for preliminary).

Another issue with notation is, e.g., X > Y and similar notations. I assume that this means that the K-index calculated from X-component is larger than the K-index calculated from the Y-component. This may belong to observatory jargon and be obvious for experts, but it must be clearly defined in the manuscript, as well as all other related notations. However, I still do not know for sure what authors mean by Y = X and Y ≠ X, which are used, e.g., in Figures 10-11. And what is meant in Sec 4.4.1 by "FMI method was also applied to TGO data with Y=X"?

Authors study in Sec 4.4.1 the validity of the X > Y assumption. I wonder why this is needed at all. Because the main aim of this research is to ensure historical homogeneity, it should be clear that the TGO method also uses the same two components (X and Y) and selects the larger of the two K-indices, just like it was done earlier. So, I am not sure if this discussion is needed. Moreover, as it was noted above, the TGO method uses the H-component. So, again, why to test the X>Y for the FMI method? This is all quite confusing and badly motivated.

In Sec 2.2.2 the authors describe the QDC choice for the TGO method. They have calculated a monthly set of daily curves averaged over the full solar cycle 22. I read this so that they did not use the actual quiet days at all. Anyway, this leads to a monthly set of so-called iron curves, which are then applied to observations made during all levels of solar activity. This is very questionable, and its motivation by ".. removing the possible problem of subjectivity when selecting quiet days.." does not weigh much in comparison to the problems that follow from this choice. Authors estimate that the solar cycle variation of the QDC amplitude is about ±10nT. Looking at Fig. 2 we can see that in winter months the typical amplitude is not more than ±5nT. Thus, if the solar cycle variation of the QDC curve was taken into account, it could have an essential effect on the K-index at certain times of the cycle. In fact this may be the reason for the difference in the TRO K-indices between the TGO and TGO' methods (the latter taking the cycle variation into account).

I also note that taking into account the cycle variation of the QDC curve is less important, not even necessary at auroral latitudes (see, e.g., Martini et al., JGR, 2016), but is important at sub-auroral and lower latitudes (like DOB), where the disturbance level is lower. Also, the effect of cycle varying amplitude may be different at auroral and lower latitudes.

It is slightly disturbing to read that the earlier version of the TGO method (TGO') did actually take the solar cycle effect into account. I wonder what made the TGO to change their method for, in my evaluation, a less appropriate version? On the other hand, it seems that authors already realize the importance of the cycle variation effect since "It is possible that it is necessary to implement less simple QDV that account for solar cycle variation of the Sq current system..".

Authors use considerable effort in trying to understand and even modify the skewed distributions of the more recent years. Alas, such modification is not needed and would even be harmful. The different distributions are due to the very weak solar activity of the last 15-20 years, when the Earth has spent most of its time in the heliospheric current sheet (Mursula et al., JGR, 2022) and the level of geomagnetic activity and storminess has considerably reduced. This has also modified the K-index distribution toward lower values. Trying to twist it back to the "normal" distribution of the more active times would be like using a medieval torturing machine. Luckily the authors seem to understand the correct explanation themselves. However, the treatment should better reflect this explanation. Authors could also check this using data from other (than Norwegian) stations.

The analysis of the K9-limits at DOB and TRO are rather futile. Because of TGO method is different in component treatment from hand-scaling, the K9-limit estimates should only be made using the FMI K-indices. Anyway, on page 13 authors should more clearly conclude that increasing the DOB K9-limit is not reasonable.

Many of the comparisons, e.g., those on the skewness of distributions, are rather vague and qualitative. Authors should aim to be able to present and use more quantitative estimates. E.g., they should calculate the values of the skewness for the different distributions and, perhaps, even test their statistical difference.

In several occasions, the text refers to a later section where the topic at hand will be explained or studied in more detail. This is a very annoying structure for the reader. The authors should try to write more directly, one topic at one time, rather than spreading one topic into several sections.

As explained, geomagnetic activity indeed maximizes at equinoxes, a fact which is known for 160 years (Sabine, 1956). The Russell-McPherron effect (projection of solar equatorial HMF field onto the geomagnetic dipole axis) contributes to this semiannual (not biannual) variation but it is not the leading contributor. The authors may find paper papers by Cliver, Lockwood, Lyatsky, Mursula, Newell, Yoshida  and many others which study the semiannual variation and their three mechanisms (two other being equinoctial (dominant) and axial mechanisms).

I think that the part on Ak time series should be dropped out completely. The treatment of Ak time series and power spectra is currently rather elementary and does not contain anything new. It should be considerably expanded and elaborated in a different publication. I find that, rather, this paper should focus on its main topic of K-indices with contents modified along the lines indicated above. It will be more solid without the added Ak annex.

Overall, it is very likely that the TGO K-indices, in their present version, cannot be used to continue the hand-scaling period indices. This is true for all stations, especially for the lower-latitude DOB. Rather, the continuation should be made with the FMI method K-indices.

Therefore, I am happy to read in the discussion that ".. FMI method... is therefore desirable to establish a repository of K-indices generated using this method for the purpose of basic research purposes". On the other hand, I cannot agree with the other statement "For operational space weather purposes, however, it (TGO method) is more than good enough as it is...". Rather, I suggest that the TGO method (maybe the TGO') should be using solar cycle variable QDC curves.

Smaller notes

There should be another Table summarizing the availability of the different types of K-indices, including the years of availability and years analyzed (if different), the method used, the component(s) used etc.

TGO' is a bad name for the "preliminary" method. Better to use, e.g., TGOp (p for preliminary).

I would favour the notation of K_X, K_Y (K_H, K_D) in order to make it clear which component is used to calculate the index.

Fig. 1 Station abbreviations would show better in other color than white, and they could then be thinner.

Table 4. Are the given K9-limits for the X component? What are they for H and Y? Please include them in the Table.

Table 2.  Correct the K=2 lower level.

Fig. 9. I see no point in including panel c here.

Figs. 12 and 13 could perhaps be joined. Enlarge Fig. 12 width.

I would prefer to use the 3-hourly time intervals as 00-02, 03-05,...with 21-23 as the last one, instead of the confusing 21-24.

Text additions like "See, e.g., .." on line 70 should be placed within parentheses. There are several of them.

The fact that the FMI K-index method is supported by IAGA has been repeated several times. Once (in Sec 2.21.) is enough.

Write consistently K-index, sub-auroral, hand-scaling, over-correction, under-correction,..

Line 58. Haldde is marked in green, not white.

Line 115. If the early K indices were not listed in the yearbook, how did they enter in the  IAGA Bulletin? I would like some clarification.

Line 132. "digital TGO data" should be more clearly defined here.

Line 296. "Two thirds" cannot be true since about 60% are perfectly matched (Fig. 7).

Line 367.  Correct: In DOB, K_HS < K_TGO, but in TRO, K_HS > K_TGO. This difference is likely due to the different effect of cycle variation on QDC amplitude at the different latitudes.

Line 425. Date is in error. The event is the same in Fig. 13b.

I find this work significant, since it offers not only a deep analysis of extensive time series of K, but also a thorough and detailed description from historical and analytical perspectives. Studies like this are valuable for dealing with series which result from concatenated measurements, something that is now widespread across numerous scientific fields dealing with long-term time series, like space weather and atmospheric and climate sciences, that have data which come from different instruments, analysts, or even changes in the measurment location.

I consider that this work can be accepted for publication in this journal with minor revisions. Below, I list some comments, followed by minor corrections I suggest regarding typographical or other minor issues.

Comments:

What is a "log-normal-esque" distribution? In particular, what does "esque" means?

In section 3.1 (Distributions during the transition from handscaling to automatic methods) I suggest complementing the analysis including a statistical index with % confidence of the comparison made in Figure 4. Maybe the Kolmogorov-Smirnov test, to test the distributions similarities, or any other of your preference.

This will complement what you mention in Line 281. There you say " We have seen that the TGO method provides a good match with the HS and FMI derived K values in the the old and new series for TRO and DOB (Fig. 4)."

In Section 5.1 (Power spectra of Daily Ak), a possibility if the series has data gaps, is to use Lomb-Scargle. This is only a suggestion.

In Figure 15, of the power spectrum, I would add the line showing the 95% confidence limit, or the red- or white-noise spectrum.

Minor comments:

Line 4: In "In t his paper," there seems to be a space in the worth "this". Please, check.

Lines 8-9: In " It becomes clear that each method "both have strengths and weaknesses." I am not sure if it should be "It becomes clear that each method have both strengths and weaknesses." Please check.

Line 58: " Halddetoppen, marked in white." Isn't it in green?

Line 79-80: " (e.g. (Nevanlinna, 2004; Nevanlinna et al., 2011)))" I think it should be "(e.g. Nevanlinna, 2004; 80 Nevanlinna et al., 2011)". Please check.

Line 116: " (e.g. Sergeyeva et al. (2021); Nevanlinna and Häkkinen (2010))." I think it should be " (e.g. Sergeyeva et al., 2021; Nevanlinna and Häkkinen, 2010)." Please check.

Line 151: " The QDVs are are the monthly averages ..." delete one "are". It should be " The QDVs are the monthly averages ..."

Figure 4 captions: "TRO (19080, 1988-1991)" should be "TRO (1980, 1988-1991)"

Line 227: "Even though the there are shifts in the distributions, ..." It should be "Even though there are shifts in the distributions, ...". That is, delete "the".

Figure 6 caption: Check this figure caption. a) corresponds to TRO and b) to DOB.

Line 434: "For these dates the fitted QDC are clearly erronous (not shown), with amplitudes up to 10000 nT.". Is 10000 nT the correct value? It seems too large. Please check.

And "erronous" should be "erroneous".

Line 443: " where KFMI = 9}." Shouldn't it be " where KFMI = {9}." Please check.

Line 488: "AK" shoulfn't it be "Ak"?

Line 534-535: " (e.g. Mayaud (1980))." I think it should be "(e.g. Mayaud, 1980)." Please, check the instructions for authors.

Line 554: " (e.g. Menvielle et al. (2011))" I think it should be " (e.g. Menvielle et al., 2011)"

Line 559: "frequncy" should be "frequency"

Line 561: "provided in (Table 4)." I think it should be "provided in Table 4."

Line 615: "have a strength" I think it should be "has a strength"