The manuscript “Over three decades, and counting, of near-surface turbulent flux measurements from the Atmospheric Radiation Measurement (ARM) user facility” by Sullivan et al. provides comprehensive documentation of surface-based turbulent flux observing at ARM. The manuscript reviews methods, history, configuration, validation, site characteristics, recommendations, available support tools, data access information, context with similar networks, etc. The manuscript provides invaluable documentation of a complex, global series of turbulent flux measurements that have been operated be ARM for more than 20 years.

I’m a user of ARM data myself, occasionally including the turbulent fluxes. While ARM data sets are known for exceptional documentation, the turbulent heat fluxes have been one of the more complex (in terms of varied application) and thus less tractable data sources provided by the organization. Therefore, this manuscript is a welcome addition to ESSD and will provide an excellent basis for researchers interested in the ARM turbulent flux products, making this one of the easiest manuscripts to judge in my career. The manuscript should be published promptly.

I have only one question (hopefully I didn't miss this). What happens to the raw, high-frequency (~10 Hz) component measurements (T,u,v,w,q)? Some researchers with specialized needs may be interested in the raw data to analyze spectral details or subsets over varying integration windows with various applications for corrections. Are the raw data available for these purposes? If they are archived, but unavailable, I recommend ARM consider releasing them with DOI (though please don’t hold up publication of this manuscript to do so). If they are not archived, I recommend ARM consider doing so in the future.

This manuscript provides a unique historical background and synthesis of methods and metadata that will be of great value to many researchers and practitioners using ARM flux data. The comparisons between instrument systems will be very useful for common problems that require combining datasets. The techniques and theoretical basis are well described. There is a good balance between providing specific details on ARM approaches and providing just enough background for context, without becoming repetitive with existing texts. The manuscript is well written, and I appreciate having had the opportunity to review it. I have only one major suggestion that is not absolutely critical but would be nice to have, and a few areas where I recommend minor edits. 

Major comments:

My only suggestion that could be considered “major” is to include some discussion and a link to data on the canopy height, if it exists. This may not be a huge concern at most of the sites due to the short-statured vegetation, but it is certainly a factor at E21. I realize that one could back out canopy height using methods such as Chu et al. 2018, but having direct estimates would be beneficial. I also recall a discussion many years ago regarding vegetation growing too close to some of the sensors (related to the more general concern of sensors placed deep within the roughness sublayer). Any clarification on that problem (if it still exists) would be helpful. 

Minor comments:

L 24: perhaps use a semicolon instead of the awkward double parentheses.

L 60: These data have been used extensively to study a range …

Since it’s not feasible to cite everyone, it may be best to use e.g. in front of the chosen citations so as not to imply that the publications are this few.

L 199: “Where w’ is the instantaneous fluctuation of the vertical wind speed component about the mean”

Rather than having to state “about the mean” after each variable in this section, you could just define the perturbation as a departure from the mean, and thereafter use the word perturbation without repeating “about the mean”.

L 265: v for versus – I believe the scientific convention is vs. - also to be consistent with line 455. Please check throughout as this occurs often.

L 441: site was excluded

L 472: postulated that differences between…

I think this section needs to be revised as it makes it sound like the differences between this study and Tang et al. 2019 may be larger than in reality. See further suggestions below.

There is little doubt that the differences at the CF site are attributable to the vegetation differences. The “postulated” wording makes this sound more dubious than may be intended, especially when paired with the text that follows. This is not to suggest that instrument differences aren’t also contributing. 

L 480: “we conclude that the differences between LE measured by the two methods (EBBR and EC) are reflective of differences in the instrument systems themselves, not solely due to environmental factor” I think you mean “differences … at the E39 site … are due to differences in the instrument systems themselves, not solely…” Since Tang et al. also compared averages over all the EBBR sites against the ECOR averages, the current wording may be taken to imply that instrument differences dominate across the entire ARM SGP domain. This was a popular viewpoint until it was encouraged to check this assumption, leading to the paper by Tang et al. 2019.

L 475: “...but was no longer significant…”

You may want to tweak this wording, as I think they did make note of non-negligible differences. From Tang et al.: “Although surface difference is the major factor contributing to the flux differences, the above results also show that instrument difference is nonneglectable.”

L 476 “However, no clear dependence of the agreement on vegetation type was observed at E39,”

I would remove the “However” because I don’t find this statement to be contradictory to Tang et al. given what is discussed next regarding the much larger separation between the EBBR and ECOR at the CF and the fact that the EBBR is in a different field entirely. Actually, their Figure 3 is not too far off from the LE and H differences shown in this manuscript.

As an aside, curiously, a much smaller difference was found between the EBBR and ECOR at Medford E32 in 2016 (see Bagley et al. 2017). Any thoughts on this? Either way, you may want to discuss this paper as it attempted something similar in comparing the two systems for the same vegetation type/fetch. Also, you may want to note that EBBR does not share exactly the same footprint as ECOR even if it is mounted at the same height and location, and this is a source of uncertainty in the comparisons. 

L 601: “This finding does not have any clear dependency on vegetation type (crop v grass).”

Here again, it reads too generally in my opinion. Yes, I agree that the analysis at E39 demonstrates the EBBR/ECOR differences, and shows minimal influence of vegetation type. Beyond E39, however, these two instrument types were distributed unequally across vegetation types within the ARM domain(s). Table B2 makes this clear. When the vegetation type (LAI) and function (GPP) are considered, as in Williams and Torn (2015), it becomes clear that there is a definite influence of the underlying vegetation in general. Many researchers want to average the data to obtain a spatial mean, and they need to know how to do this in a way that both represents the mixture of land surface characteristics and deals with instrument differences. My concern with the current wording is that some researchers will choose one instrument type or the other, in order to avoid dealing with instrument bias, and then proceed to average over only ECOR or only EBBR sites, thinking that they are getting a good spatial representation. Unfortunately, this approach was used many times in the past. I recommend adding a statement to the effect that, when aggregating data spatially, one needs to factor in both the instrumental differences as well as differences in underlying land surface characteristics. I believe that this manuscript does a nice job of providing the information needed for individual PIs to combine datasets as they see fit for their purposes.

Chu, H., Baldocchi, D. D., Poindexter, C., Abraha, M., Desai, A. R., Bohrer, G., et al. (2018). Temporal dynamics of aerodynamic canopy height derived from eddy covariance momentum flux data across North American flux networks. Geophysical Research Letters, 45, 9275–9287. https://doi.org/10.1029/2018GL079306

Bagley, J. E., L. M. Kueppers, D. P. Billesbach, I. N. Williams, S. C. Biraud, and M. S. Torn (2017), The influence of land cover on surface energy partitioning and evaporative fraction regimes in the U.S. Southern Great Plains, J. Geophys. Res. Atmos., 122, 5793–5807, doi:10.1002/2017JD026740.

Williams, I. N., and M. S. Torn (2015), Vegetation controls on surface heat flux partitioning, and land-atmosphere coupling, Geophys. Res. Lett., 42, 9416–9424, doi:10.1002/2015GL066305.