The manuscript presents a worldmap of patterns of enhanced NO2 over oceans that can be attributed to ship tracks or oil rigs, based on high-pass filtered NO2 measurements from TROPOMI. The paper is well generally well written, and the resulting ship tracks are impressive by the presented detail.

I have the following general concerns:

- The paper presents patterns in a rather qualitative way (tSCDs), while in the title as well as in some instances in the text it sounds as if the study aims to quantify ship emissions. This should be avoided.

- The methodology should be clarified and the individual retrieval/filtering steps should be illustrated for real measurements.

- The comparison to CAMS VCDs is problematic, as this dataset has not been high-pass filtered. This should be modified, or discussed properly.

- The outlook/next steps should be discussed in more depth. Obviously, it would be desirable to quantify NOx emissions for the detected ship tracks. How to do this, which problems have to be solved? (impact on filter settings, AMFs, ...)

Detailed comments:

Title: The title is misleading, as it suggests that this study is about quantifying ship emissions. Please consider modifying this, especially in the manuscript, for instance in line 149: I would expect to read a number at the end of the section with such a headline. 

A statement like that in line 382 should already be made in the abstract.

Line 4, line 49: This TROPOMI pixel size only holds for nadir. Towards the swath edges, pixels are considerably larger.

Line 90: I agree that the tSCD allows the detection of ship NOx without bias from a-priori knowledge used for the AKs in the operational product. However, as soon as NOx emissions should be quantified, the tSCDs need to be converted to tVCDs by an (appropriate) AMF.

Please clarify.

Lines 93-104: I think a short modification sentence would be helpful: The tSCD is provided in the operational product (or could easily be calculated by subtracting V_strat*AMF_strat from total SCD), but this quantity turned out to be affected by a simplification in the operational processor: ...

Line 106: What does "approximately" mean?

Line 109: How far is this a "saturation"? What is saturated?

Filtering: The iterative filter approach seems complex, and Fig. 1 does not really help. I would propose to have a sketch showing the effect of the individual steps on real data instead.

Fig. 2 shows the result of the iterative filter process. I do not understand how the values here can exceed 4e13, when in a previous step everything above 3e13 has been skipped. Please clarify.

Line 149: Even with stating the disadvantages of applying filters, one might think that all the discussed filters are applied below. A statement should be added that, as default, they are *not*, with reference to later discussion why not.

Line 157: except the model used for assimilating strat. NO2...

Fig. 3: The map of resulting ship tracks is quite impressive, and the benefit of 1° over 0.25° quite prominent. However, this is a result of the high-pass filter, and this will inevitably cause *negative* tSCDs next to the shipping lines. This would have to be taken into account for quantification of emissions. I thus think that the figures should also show the negative values to create awareness for this effect - this might not look as nice, but more honest.

CAMS comparison:

Are the CAMS tVCDs high-pass filtered as well? If not, they should for a more meaningful comparison.

Line 421: and avoiding artificial ship tracks just introduced by the a-priori profiles?

Fig. C1: This comparison is not appropriate, since the TROPOMI data has been high-pass filtered, but the CAMS data not. At least this aspect needs to be clearly discussed.

A statement on data availability is missing. I would encourage the authors to make their results available on a data repository so that future studies could reference this dataset with a doi.