Dear authors,

thank you for your response to the reviewers' comments and suggestion. Please go ahead and revise your manuscript accordingly.

Sincerely,
Kai Schulz

Dear authors,

your revised manuscript has been gone through a second round of reviews. While the first referee deems your amendments adequate, the second referee has put forward a number of additional suggestions (see attached pdf) which should be incorporated into the final version.

I suggest that you are more clear about the uncertainties of your method (mentioned in the abstract), i.e. drift, accuracy, suitability in environments with lower carbonate chemistry fluctuations and also discuss means to address them, e.g. frequent (at least daily) calibration against discrete high accuracy pH measurements (what about sensor drift at night in comparison to during the day?).

Sincerely,
Kai Schulz


Furthermore, the second referee raised the following points:

REVIEWER

Certainly the authors have made many changes that do improve the manuscript. However, I do not feel entirely comfortable with the assertion in the authors' response: “While accuracy was an important factor, our intention, with proof of concept data, was to demonstrate that this technique is capable of showing clear patterns of variability.

First, the abstract emphasizes “to examine diel fluctuations in seawater carbonate chemistry”, not sensor “proof of concept”.

Certainly the design is feasible, and the variability shown is plausible. I am however uneasy with the magnitude of pH adjustment that was carried out (lines 171-172) of 0.0136 pH units per day (looking at Figure 5, this seems to equate to the total data amplitude!!). (As noted, the data seem plausible, so adjustment may well be mostly right)

I also wonder at the likely significance of the various figures of merit quoted for this system:

resolution 0.0028 pH units (inferred from the manufacturer’s specification for the phase angle resolution, and (I assume) the “calibration curve” in phosphate buffer mixtures.

precision 0.022 pH units (standard deviation of 3 h of measurements in a seawater sample). This seems a bit high. It is not clear that the sample remained stable for 3 h, and - of course - there would be about 0.004 pH units of drift (see above) in 3 h. (see also my off-the-cuff estimates of repeatability below)

accuracy data are quoted as pHT (based on adjusting for drift to spectrophotometric measurements with a likely uncertainty of about 0.015 or so (my estimate not theirs)

A simple examination of figures 4 and 5 suggests a repeatability of measurement (based on the differences between consecutive measurements - i.e the width of the band) of around 0.004 in pH in April, and just below 0.010 in pH in December. I think I would encourage the authors to calculate this parameter properly,** as it seems a more meaningful one than their 0.022 above. However, I wonder if the difference is due to different sensors, different measurement electronics, or some other factor?

If one is looking to measure fluctuations, it would be good to have a reasonable estimate of how well a peak to peak difference in pH can be estimated (related to the ratio in hydrogen ion concentrations). I do not see such information here. Also, the drift is considerable, and assumed to be uniform over an extended period (>1 week), so there is a possibility for additional error to come from this hidden contribution if it is not uniform.

Also (a minor point) the salinity values are probably not very accurate, this is not really improved by using ppt as I suspect this implies a conductivity equivalent to that value of ppt in NaCl solutions.



** The approach I would take is to recognize that a repeatability can be calculated from a set of duplicate measurements (with degrees of freedom equal to the number of pairs), and assuming that two consecutive measurements can be treated approximately as such a duplicate pair, implying insignificant real change in pH between consecutive measurements.

Then the repeatability (standard deviation) can be calculated from the following expression (and based on that for a pooled standard deviation) where d is the difference between the pair, for k pairs of data.

sr=( sum(d_i^2) / 2k)^1/2

Dear authors,

thank you for addressing and clarifying the remaining issues. Your manuscript will be accepted for publication in Biogeosciences.

Sincerely,
Kai Schulz