I thank the authors for responding to my previous remarks in their revised manuscript. Largely the changes that they have made are satisfactory. They have fixed equation (2), using variances instead of standard deviations. They have also changed their figures to show variances rather than standard deviations.
Furthermore, they now go beyond just mentioning the mean state, touching on how results change when that is included in the analysis. Unfortunately, their demonstration of the importance of the mean state comes too late and is too brief. Appearing only as a glimpse in the middle of the Discussion section, many readers may well miss it. I think it should be expanded upon further and presented in the first part of the Results section. Right now it seems to come as only an afterthought in the Discussion, a peculiar place given that it is the change in the mean state that is the dominant factor controlling the change in actual extreme events for both [H+] and ΩA.
In an attempt to fix previous confusion, another reviewer suggested that the authors should generally replace "extreme event" with "extreme variability event", and the authors have complied. Although well intentioned, when I read the revised manuscript, the ambiguity of this term, or rather what the authors mean by it, seems to stand out. While "variability" refers to an oscillation over a given period involving both highs and lows, "extreme event" refers to the time or times at which a high or low threshold is surpassed.
A search with Google Scholar reveals no precedent for use of "extreme variability event" in climate science. Although it has been used in astrophysics in several published papers, the meaning is different. It is only used to refer to events with high variability. It is never used, in any field as far as I can tell, to refer to a maximum or a minimum generated by that variability crossing a upper or lower limit (after removal of the mean), i.e., the meaning intended by the authors. Due to this confusion and lack of precedent, that term should be avoided here. More careful wording is still needed.
Unfortunately to fix this ambiguity is not so simple. Sentences and sometimes full paragraphs will need to be revised on a case-by-case basis. In the specific comments below, I give a couple of examples of how the authors could rewrite a sentence (L33-34) and a paragraph (L346-352) to avoid this confusing 3-word term. Other specific comments also address how to improve some related passages, while elsewhere I leave it to the authors make the relevant improvements.
Overall then, I recommend that the revised manuscript be further improved by expanding on the section concerning relative importance of the changing mean state, moving that expanded subsection up to the beginning of the Results section, and clarifying all passages where the authors currently used "extreme variability event" or similar ambiguous terminology (see specific comments).
- In some places the authors still use "daily" instead of "daily mean" (e.g., L133, 134, 159)
- replace "GFDL ESM2M" with "the GFDL ESM2M model". Added clarity is needed at least for non-modelers. Readers tend to skip around in a paper and may well miss the original definition.
I am confused by the title. The use of "extremes" appears misleading because the paper does not actually focus on extreme events.
L1: I'd suggest to change "extremely" to "relatively". The sentence then becomes more subtle and avoids word repetition.
L2-3: For a clearer and less verbose sentence, I'd suggest to change the last half to "not only due to changes in the long-term mean but also changes in short-term variability."
L5: The phrase "changes in variability on variability-driven ocean acidity extreme events" is not clear. A common but wrong interpretation from readers could be how variability contributes to extreme events in the context of the dominant role of the changing mean state. The term "variability-driven ... extreme events" does not clearly indicate that the changes in the mean state are ignored and that the focus is actually not on extreme events but rather changes in extremes generated by variability alone.
L8: same problem as just above with the term "variabillity driven extreme events"
L10: ambiguous use of "extreme event".
L12: change "variability extremes" to something like "the associated increased maxima from that variability alone"
L13-14: This sentence regarding OmegaA is open to misinterpretation due to the ambiguous term "variability driven extremes" and the opposing effects from changes in variability and changes in mean state on actual OmegaA extreme events. As space seems to be lacking to get into enough detail, I would suggest simply to delete this sentence.
L15: change "will" to "may"
L23: change "formation of [HCO3-] from [CO32-]" to "conversion of [CO32-] to [HCO3-]"
* What qualifies as "extremely low"? Would it not be better to say "much lower than usual" because it is the relative difference that matters?
* Does not variability also lead to extremely high values of pH and OmegaA not only low values? As phrased, the statement is one-sided.
The problem is the term "extreme variability events", which implies changes in both highs and lows. While variability refers to oscillation over a given period, "extreme" and "event" refer to a moment in time when a threshold is surpassed (as mentioned in the General Comments).
To remedy the problem, I would suggest to change the following sentence from
"Superimposed onto the long-term decadal- to centennial-scale ocean acidification trend are short-term extreme variability events on daily to monthly timescales, during which ocean pH and/or Ω are extremely low (...). These events ..."
to something like
"Superimposed onto the long-term trend are lows and highs in pH and ΩA that will be modified as short-term variability changes (...). Lows from extreme variability ..."
L42: This sentence does not seem to make sense since "extreme variability events" are anomalies relative to a long-term mean and mean pH conditions are not anomalies. Perhaps the authors mean "changes in mean pH"? In any case, "extreme variability events" is ambiguous and should be avoided.
L46: change "variability extremes" to "variability". That is, "variability extremes" refers to when variablity is high or low but not the min or max generated by that variability (anomaly relative to the mean).
L50: delete "events"
L51: "saturation" is when OmegaA = 1. "Supersaturation" is the word used when OmegaA > 1.
* insert "ion" after "carbonate"
* "several days of aragonite undersaturation" is unclear and open to misinterpretation. Change to "several days of being exposed to waters which are undersaturated with respect to aragonite".
- Please tone it down a bit by changing "of critical importance" to "important"
- suggest to change "and the changes therein" to "and how that will change"
L75: change "impose changes in" to "affect"
2.1 MODEL & EXPERIMENTAL DESIGN
L120: delete "long"
2.2 ANALYSIS METHODS
The word "METHODS" seems unnecessary since this subsection is already in the METHODS section.
2.2.1 EXTREME EVENT DEFINITION AND CHARACTERIZATION
L132: "sulfate" ions? The authors' defintion of the [H+] total scale is wrong.
* change "daily" to "daily mean"
* change from "a one-in-a-hundred days event" to "occurring once every 100 days."
* I do not buy this statement, as written, as justification for using relative thresholds over absolute threshold. Absolute thresholds can be determined as a function of a regionally (or grid-cell) varying baseline (preindustrial conditions here).
* "statistical properties" is vague.
L136-138: I don't follow. Biases in simulated variables also alter the definiton of model-defined absolute thresholds if they are regionally varying.
L138: "extreme events" is ambiguous.
L135-138: It seems that the authors are trying to use the last 4 lines of this paragraph to justify their choice of studying only the effects from variability and ignoring the changes in the mean state, but the text is not clear nor convincing. Moreover, the text only exposes the advantages. The other side of the story is that an analysis of the relative thresholds has little to do with actual "exteme events" (the authors' own term in the last sentence) when changes in the mean state are dominant (the case here for [H+] and ΩA).
2.2.3 TAYLOR DECONVOLUTION METHOD TO IDENTIFY MECHANISTIC CONTROLS OF [H+] AND ΩA VARIABILITY CHANGES
This 13-word subsection title is too detailed. Please change it to something simpler such as "Taylor expansion". The first sentence of this section provides the more detailed complementary information.
L198: twice in this sentence, we see "drivers" modifying a noun. It should actually be "drivers' " because in both cases it is a plural possessive.
L199: The sentence "We do so by calculating the full Taylor series ... up to the fifth order." will confuse readers for 2 reasons:
* no Taylor series is "full" (in practice there is always truncation of higher order terms)
* Mentioning an "order" in a sentence about a Taylor series should refer to the order of the Taylor series. But the authors' Taylor series equation is only 1st order. It does not even include terms with 2nd derivatives as needed for a 2nd order Taylor series. After looking at Appendix C, I understand a little more about what the authors might be trying to say, but it would be necessary to separate the two different meanings for "order" in 2 separate sentences. Better yet, avoid all confusion by not even mentioning fifth order here. It is not critical to the text, at least not in this section.
L200-204: Appendix C should be mentioned earlier. Its first mention does not come until ten lines later (L214), so readers will struggle trying to make sense out of the authors' introduction of the 4 new terms, which are not so simple after all.
L207: Eq 3 is verbose. Eliminating the H+ subscript would be an improvement and make it more general. Making that change would mean that the authors should also generalize parts of the previous text, including for example changing the end of the sentence on line 201 to "overall change in variance for each of H+ and ΩA (∆sσ2)"
L208-210: This sentence seems out of place. First it mentions Figs 8-10 too early (before Fig. 3 is introduced). More importantly, the concern turns out to be minor since the authors show that the grey and black lines generally agree well. I suggest to move the idea of the agreement to the the caption of Fig. 8 or to the text that discusses the results from those figures.
L212-214: The mathematical terms are becoming even more verbose. Again, removing the H+ subscript would help.
2.3 MODEL EVALUATION
L227: "co2sys" should be capitalized (all letters).
L236: simplify the parenthetical statement to just "(Appendix D)"
Fig 3 (contents): The Landschutzer data product does not extend all the way up to 90°N. Please modify the label of the northernmost band accordingly. Also the Southern Ocean does not extend all the way down to 90°S.
Fig 3 (caption): Change the cumbersome phrase ". (c,d) The same as (a,b), but for ΩA." to "along with the same for c) data-based ΩA and d) simulated ΩA." Other figure captions also have the same type of cumbersome phrase and should be improved in the same way.
L249: insert "seasonal amplitude of" before "ΩA".
L257: delete "rather".
L264: change "was" to "were"
L262-265: Four "we" on 4 consecutive lines is a bit too much.
3.1 GLOBAL CHANGES IN OCEAN ACIDITY VARIABILITY EXTREMES
* panel a title: "Yearly extreme days"
* caption: "frequency"
* caption: "number of extreme event days per year"
Fig. 5: "Yearly OmegaA Extreme Days"
Line 297: delete "only"
line 303: change "long" to "longer"
line 307: insert "period" after "historical" and insert "during" between "and" & "the"
line 310: delete "It should be noted that".
3.2 REGIONAL CHANGES IN OCEAN ACIDITY VARIABILITY EXTREMES
L314-327: This paragraph is weak because it only quantifies absolute changes. We also need to know the numbers for what were the percent changes relative to the preinidustrial reference.
L342: change "similar as" to "similar to"
L340-345: It would be nice to remind readers that you are discussing results for [H+].
(2) L346-352: This paragraph is particularly confusing because (a) it relies on the ambiguous term "extreme variability event" and (b) its conclusions (an "improvement" with time due to declining variability alone) are opposite to those for a real "extreme event" (a "worsening" with time, i.e., for ΩA where the long-term declining mean dominates the signal). Here is the authors' text:
"While the decline in mean ΩA generally leads to lower values in ΩA, extreme variability events in ΩA are projected to become less frequent throughout most of the ocean (89% of surface area under RCP8.5 at the end of the 21st century; Figure 5c). In many regions, extreme variability events in ΩA are projected to disappear by 2081-2100 under the RCP8.5 scenario (grey regions in Figure 5c). However, the frequency of surface ΩA variability extremes is projected to increase by 10 or more days per year in the subtropical gyres, especially in the western parts of the subtropical gyres. At depth, no extreme variability events are projected for most of the ocean during 2081-2100 under RCP8.5 (Figure 5d)."
I would suggest that the authors try to lead readers more by the hand while avoiding the term "extreme variability event". As food for thought, here is an attempt to make this paragraph clearer:
"Although the long-term mean of surface ΩA declines, so generally does its projected variability. Thus after removing the mean, there is also a decline in the number of days per year when the variations lead to levels that are below the reference threshold (1st percentile from the preindustrial variability distribution with the mean removed). That decline in frequency is evident over 89% of the ocean; by 2081-2100 under RCP8.5 the low reference threshold is never reached over most of the ocean (grey regions in Figure 5c). Conversely, in the subtropical gyres the low threshold is crossed with increasing frequency, reaching 10 or more days per year during 2081-2100 under RCP8.5. Simultaneously, at 200 m, the corresponding low reference threshold is never reached over most of the ocean (Figure 5d)."
3.3 DECOMPOSING [H+ ] VARIABILITY CHANGES INTO INTERANNUAL, SEASONAL, AND SUBANNUAL VARIABILITY CHANGES
This subsection title is verbose. It should be changed to something like "Decomposition of temporal variability"
L355: change "We therefore decompose" to "Thus we decomposed"
L356: change "overall" to "generally"
L360: "preindustrial" is a period, while "the end of this century" is a point in time. "Preindustrial" is also an adjective and should be modifying something? This issue seems a common mistake in this manuscript.
- "extreme variability event days" is a confusing term.
- Do the authors really mean "extreme events"?
L360-367: It would be easier to understand if the authors would just refer to increasing variability or variance rather than getting into the messy "extreme" terms, which they use currently. This paragraph only discusses Fig. 7, and that figure only shows the variance. The word "extreme" seems entirely unnecessary throughout this paragraph.
For the same reasons as mentioned just above, I would also suggest the following:
* L370: change "variability extremes" to "variability"
* L374 and 378: change "extreme variability events" to "variability"
* L377: delete "extremes"
* L373: change "types of variability" to "time components of variability"
* L377: change "variability types" to "time components of variability"
3.4 DRIVERS OF [H+] AND ΩA VARIABILITY CHANGES
L390: delete "northern and southern". As no other "high latitudes" exist other than in the north and south. Thus, these 3 words are unnecessary.
L394: insert "The" before "GFDL" and insert "model" after "ESM2M"
* insert "to" after "leads"
* change "very" to "more"
L397: "This" what?
L403: "that following" is confusing. Changing those 2 words to "increases" would be better if that is what is meant. Unclear.
L407: "would else"? What does that mean?
L409: simplify "a large part" to "much"
L410: change "golden" to "gold"
Figure 8: Panel (h), the zonal mean plot, is very important but also very busy. The authors try to plot too much information, 8 lines, which is even more confusing because of some overlap (green and gold). I would suggest to make a separate zonal mean figure, where a first panel shows only the first 6 lines, and subsequent panels break down contributions to each term (total, s, sigma, s-sigma, rho) from the different components (CT, AT, T, and S). This would also help the authors to clarify their text on L403-412, which currently (without such added information) seems to suffer from the common a priori concept that only CT matters. If the authors feel that adding a separate zonal mean figure, would result in too many figures, I would suggest that the suggested zonal mean figure should be kept in the main body of the paper and that the maps should then be moved to an Appendix or Supplement.
Figure 9: Panel (h) the same comments made just above concerning Fig. 8 also apply to Fig. 9. In addition, it is hard to distinguish the 2 blue colored lines, especially because they are generally so close together. Furthermore, it is bothersome that the choice of axis limits cuts off the largest variability seen across large zonal bands.
Figure 10: The line color for each of the different components should be made entirely consistent with those used in Figs. 8 and 9. Currently they are not.
L434: change "similar relative" to "secondary". Otherwise, the authors are imposing on the readers to recall what was said about [H+], which they may not have read or may have forgotten amidst a wealth of other information.
DISCUSSION AND CONCLUSIONS
L436: I would recommend to delete "and conclusions" in this section title. By default, the Discussion includes the conclusions if there is no separate Conclusions section. Actually, modern science readers seem to often prefer to have a separate Conclusions section because with more and more papers to read, the tendency is to read only parts of a paper.
L437: Please rephrase without using the confusing term "extreme variability events".
L439: No, the manuscript does not focus on "extreme events" unlike what the authors state here. Rather it focuses on periods where variability is high. More precisely, after removing the dominant contribution from the change in the mean state it focuses on the resulting upper value (for H+) or the lower value (for OmegaA).
L440: suggest change "extreme variability events" to "extremes in variability"
L442: change "event characteristics" to "variability".
L443-444: change "extreme variability events" to "variability"
* change "Extreme variability events" to "Extremes in variability". It would be even clearer to replace this sentence with "Variability of OmegaA is projected to decline in the future."
* change "it is because" to "The reason is that"
- change "significantly add to the reduced occurence of ΩA variability extremes" to "further reduce ΩA variability"
- To avoid ambiguity, "significantly" should not be used except when discussing statistical results. Even then it should be elaborated upon, with something along the lines of "statistically significant".
L450: use of "extreme variability events" should be avoided. A more precise first sentence would be "Here we analyzed how extremes driven only by variability change, i.e., after removing the long-term mean."
L459: insert "model" before "projects"
L466: delete "very"
L469-470: I do not understand this sentence starting with "In other words". Either delete it or be clearer about what is being referring to. The authors have talked about 2 criteria, but it is not stated which criterion each of the numbers mentioned here is referring to.
Figure 13: (caption) delete "temporal"
L477: It is strange to cite Bednarsek for the meaning of ΩA = 1. The meaning was defined well before that paper.
L479: change "the ones" to "those"
L482: insert "the" before "surface"
L492-493: The authors state the following:
"Here we show that the changes in the seasonal cycle of [H+] translate into large increases in short-term extreme acidity events, at surface as well as at 200 m."
But where do the authors actually show this? From looking at their results, my impression is that the change in the mean state is by far the main reason why the extreme events in [H+] increase, and that the effect from changes in the seasonal cycle is of second order. Please be quantitative.
L495: "extreme variability events"? What is meant here? Rephrase.
* What is meant by "temporal"?
* Can the authors be more precise about what is meant by critical? Have not they shown that most of the variability will be captured with only monthly mean output?
L502, L503, 521, ...
- replace occurrences of "GFDL ESM2M" with "the GFDL ESM2M model". Some readers will not be modelers. Any added clarity is worthwhile here. Acronyms are confusing and some readers will reach the Discussion before seeing the authors' definion of GFDL ESM2.
L508, 509, 527, etc: Please rephrase around all occurrences of "extreme variability events"
L512: Break this sentence into 2 sentences, splitting it just after the 2nd "model".
L559: I would recommend to avoid footnotes entirely. They are generally frowned upon by journal editors and not even allowed in many journals for good reason.
In this case, we see complications when the footnote is split across pages, half of it appearing under another Appendix, and it does not get the Equations numbers correct. An easy solution is just to move the text and equations that are currently in the footnote to the body of the appropriate Appendix to which the footnote refers to. Adding a separate paragraph as an aside is not a problem.
L578: The authors will confuse readers by saying that their Taylor series is 5th order. It includes only 1st derivative and is thus a 1st order Taylor expansion. Saying things like "its Taylor series has five orders" is ambiguous. That should just be deleted. The order of a differential equation has a particular meaning. What the authors mean by order is not carefully defined. Perhaps they can find another term or carefully define what is meant.
L631: I think that this type of Data availability statement is no longer adequate for the journal Biogeosciences. My recent experience is that some model output will actually need to be made available up front without the requirement to first come back to the authors. The authors might want to check the latest author guidelines and consult with the editor.
L635: "significantly" is ambiguous.
Table A1: the table title should be placed at the top of the table. It should be short, unlike a figure caption.
Figure A1: "extreme [H+] days" is ambiguous as is "extreme [H+] variability events".
=== END ===