Basically I find the paper worth publishing, in which a new python tool MOSFiT for magnetic observatory data analysis is introduced. As the source program has been made open, it can be used by observers and researchers widely for different purposes, not just for secular variation analysis but for data processing as well. To improve the manuscript before publishing, I would give a suggestion as below, for which some further computation (but not tough) is needed to supplement their quantitative results. I would like the authors to consider it and make a necessary revision.  

On top of its useful function for jerk detection, MOSFiT is featured with various methods for external field mitigation. To characterize the different methods more in detail, I suggest that quantitative comparison of SV misfits derived with all those methods be made in Section 4.1, where only two cases (CHAOS-7 corrected and uncorrected SVs) are compared in the manuscript. Analyses for KP, QD and NT method, or even for a combination of their use with the CHAOS-7 method, can also be made to complement Table 3 with their outcomes. Furthermore, Figure 9 and the text for its explanation in Section 5 may be moved to Section 4.1, as a qualitative illustration to support the quantitative analyses, which can be referenced from Section 5 now dedicated solely for jerk discussion.  (Note there is redundancy in L229-231 for describing Fig.9. Some have already been stated previously in L225- 229)

It is often observed that K-indices are rather small in recovery phase of storms, when the ring currents remain, still lowering X-component level. Therefore I imagine, a combination of the KP-method and CHOAS-7 method (or otherwise a selection with Dst-index) would be even more effective for excluding external fields from data.

Typos:

L15 variesranges

L17 Abrupt changes in SV ‘trend’ (as described properly in L145)

L118 can be can be

L193 latest the end

S. Asari

GENERAL COMMENTS

The authors present their experiences related to the analysis of jerks, which are unpredictable events that suddenly speed up secular variations (SV) of the Earth's magnetic field.

Identifying jerks is not an easy task. The authors of this article undertook the difficult task of creating generally available software that performs such a task. The software, called MOSFIT (Magnetic Observatories and Stations Filtering Tool), works with geomagnetic data of Definitive or Quasi-Definitive status in IAGA-2002 format, which are easily available, for example, on the WDC (World Data Center) Edinburgh server. This applies to both INTERMAGNET observatory data and other observatories.

The MOSFIT package can also be used for quality control of data from geomagnetic observatories. The software has filtering, selection, and data visualization capabilities. MOSFIT was written in Python language, which is becoming increasingly popular. The authors provide a link to the software, as well as a guide to facilitate the use of the software package.

In the first chapters, the authors introduce the reader well to the subject of jerks and the problems that arise when detecting them. This is about isolating SV from final recordings of the entire geomagnetic field that also contain the influence of the external field. This is really a big challenge, because not all jerks are very expressive.

In order to isolate jerks, the package offers Hampel filtering as the first step. In the next stage, the contribution of the external field can be reduced by selecting data according to local midnight, Kp geomagnetic indices, quiet days, disturbed days, and by subtracting the magnetospheric field based on the CHAOS-7 model.

It is worth emphasizing that the package offers the possibility of visualization at every stage of work.

The authors presented the detection of jerks from 2007, 2011, and 2014 and compared the results with an analysis performed and published by another researcher. Generally, it can be stated that the consistency of detection was very good, both in terms of the time of occurrence of the jerk and its amplitude.

In summary, I am happy to recommend the manuscript for publication after making minor corrections listed below, in sections SPECIFIC COMMENTS and TECHNICAL, LANGUAGE AND OTHER REMARKS.

SPECIFIC COMMENTS

In the bibliography (References), it would be good to add DOI (Digital Object Identifier) identifiers if they are available. DOI identifiers are important because they add credibility to the source and make it easier to access publications. Currently, providing DOI’s is basically a standard in bibliographies.

TECHNICAL, LANGUAGE AND OTHER REMARKS

Lines 1,7,44,155,236     The word python should be rather capitalized in this context.

Line 15                      There is missing space between “varies” and “ranges”

Line 39                      Rather should be “close-to-final” (missing ”-“)

Line 58                      “computacional” -> “computational”

Line 126                    There is missing space between “time” and “series”

Lines 174,183          “table” -> “Table”

Line 232                    “Fig 8” -> “Fig. 8” (missing dot)

Line 236                    Should be rather “one-minute”

Line 282                    Change the capitalisation: Journal of the American Statistical

                                    Association

Line 310                    Change the captalisation: Generalized Hampel Filters