see attached pdf

This manuscript uses variability length scales to characterise dimethylsulfide (DMS) distributions in the global ocean. There have been many attempts to parametrise/predict DMS concentrations, but often they are insufficient for describing either specific events or global values. DMS in the ocean is controlled by many biotic and abiotic factors, which makes the task difficult. However, since sulfur via DMS emitted to the atmosphere is abundant and plays an important role in aerosol and cloud formation over the ocean, there is great interest in understanding DMS distributions in space and time. It is also of great interest to understand how dissolved DMS and its subsequent flux to the atmosphere are altered through climate/environmental change. The method outlined here seems to be robust and relatively universal – can be applied to global measurements. The finding that there is a characteristic variability length scale range for DMS is interesting and useful. The manuscript is generally well written and clear and should be accepted for publication after the following comments have been sufficiently addressed.

Specific comments:

• General – all data from the PMEL database that is used should be referenced (if there are references in the database). If there are no references, then the acknowledgement (as has been done already by the authors) is appropriate. This is standard practice (e.g., SOCAT data policy, which can be found here:

https://www.socat.info/wp-content/uploads/2020/06/Data_Use_Statement-for- 

SOCAT_v2020.pdf).

• Line 53 - reference typo
• Line 67 - first sentence should reference works that are appropriate, as it says few studies (implying there have been studies). Perhaps appropriate is Zindler et al. (2014, https://doi.org/10.1002/2014GL059547) as it uses high resolution data and investigates the role of eddies (although not sub-mesoscale, I suppose, but regional). Also, was an effort made to check the literature for appropriate datasets that were not yet input to the PMEL database? Seems like there are several more that could be used.
• Line 105 – Perhaps the Saltzman et al. (2009, 10.5194/os-5-537-2009) paper should also be referenced here?
• Line 163 – It might be useful to give a brief explanation as to why only these parameters were chosen for analysis. This point comes up again in the section on limitations. It might be useful in that section to learn some way to incorporate important missing parameters in this type of analysis in the future.
• Section 4.4 – This paragraph is a bit unclear to me. I do not understand why the first sentence is used. It seems completely disconnected from the next few sentences. Also, the sentence starting on line 320 is not so clear – what is the powerful tool provided? I think this is just a wording problem. Finally, the last two sentences make wishy-washy arguments (is this why there is a question mark in the subheading?). I guess it is no big surprise that physical mixing and biological activity drive DMS distributions and to say that DMS spatial variability is partially explained by these factors seems too simple (and a bit like we are starting here from zero – the DMS community knows the general drivers of DMS distributions, but can’t predict them). The next sentence says that the significant correlations presented here were again found by averaging – which is exactly the problem underscored on line 323. Is there an attempt here at solving something? Are the authors trying to say that the parametrisations can be improved by find the right combination of parameters exhibiting sub-mesoscale variability?  Or is there some other point to this paragraph?