Using feature-based verification methods to explore the spatial and temporal characteristics of the 2019 chlorophyll-<i>a</i> bloom season  in a model of the European Northwest Shelf

Mittermaier, Marion; North, Rachel; Maksymczuk, Jan; Pequignet, Christine; Ford, David

doi:https://doi.org/10.5194/os-17-1527-2021

Articles | Volume 17, issue 6

https://doi.org/10.5194/os-17-1527-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/os-17-1527-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 17, issue 6

Research article

|

01 Nov 2021

Research article |

| 01 Nov 2021

Using feature-based verification methods to explore the spatial and temporal characteristics of the 2019 chlorophyll-a bloom season in a model of the European Northwest Shelf

Marion Mittermaier, Rachel North, Jan Maksymczuk, Christine Pequignet, and David Ford

Download

Final revised paper (published on 01 Nov 2021)
Preprint (discussion started on 02 Nov 2020)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review', Anonymous Referee #1, 11 Dec 2020
- AC1: 'Reply on RC1', Marion Mittermaier, 19 Mar 2021
RC2: 'Reviewer comments', Anonymous Referee #2, 20 Dec 2020
- AC1: 'Reply on RC1', Marion Mittermaier, 19 Mar 2021
EC1: 'Major manuscript revisions', Andrew Moore, 21 Dec 2020
- AC1: 'Reply on RC1', Marion Mittermaier, 19 Mar 2021

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Marion Mittermaier on behalf of the Authors (19 Mar 2021) Author's response Manuscript

ED: Referee Nomination & Report Request started (31 Mar 2021) by Andrew Moore

RR by Anonymous Referee #1 (09 Apr 2021)

Suggestions for revision or reasons for rejection

The manuscript has undergone considerable revisions and has become much easier to follow. The comparison of AMM7v11, instead of AMM7v8, is helpful and allows the manuscript to be more focused on helpful examples.

# general comments

The removal of AMM7v8 results and the addition of more illustrative examples has in my opinion improved the manuscript considerably. It is much more readable and the examples help to understand the method and interpret the results.

The example shown in Fig. 5 is very helpful but could be expanded easily to be of even more use. To better understand the effect of the MODE parameters that were chosen, it would be nice to show the matched objects, for example by using the same color for objects that match, by connecting them with lines, or by adding numbered markers that also show the order in which they are matched. As a reader, I'd be interested, for example, if AMM7v11 objects 7 and 12 (off the coasts of Denmark and Germany) remain unmatched or are matched with the big L4 object 10. Adding information about the matching will help readers understand the results in the following sections 4.3 which focuses on the fraction of matched objects and problems when matching objects.

Another example of where an example could be expanded is Fig. 11: It would be nice to use the same colors as in Fig. 10 here, so that the bloom events can be related to each other. Lines could connect dots of the same colors, to further emphasize that they are associate with the same bloom event. Going one step further, Fig. 10 and 11 can be merged. And once colored, Fig. 11 appears to contain all the information contained in Fig. 10a, and 10a could be removed.

Several smaller implementation aspects of the study remain vague or are not mentioned. I have pointed out several instances in my specific comments below, one example is the log-transformation of the chl-a field. The manuscript includes a somewhat lengthy description of the choice of thresholds and that they are equally spaced in log-space, just to translate them back into linear space and discard all but 3 of them for this study. Meanwhile, an important implementation aspect concerning the log-transformation -- if smoothing is performed in log-space (I presume so but I am not sure) -- is not mentioned at all.

# specific comments

l 19: "By contrast the AMM7v11 analyses produces more bloom objects in deeper Atlantic waters, which are not detected by the satellite product.": Not if sure if this is intended but this formulation makes it sound like the model is correctly producing a chlorophyll bloom while the satellite (incorrectly) fails to detect it.

l 72: "that is": It's not clear if "that" is referring to blooms or "oceanic primary production". I assume it is the latter and would suggest to remove the "that is" and replace it with a ",".

l 77: While there is definitely a link between Chl-a and phytoplankton biomass, many of the biogeochemical models often have separate variables for both phytoplankton biomass and Chl-a. So using model Chl-a instead of model phytoplankton biomass to infer phytoplankton biomass is not a common use case. But why mention biomass here at all, I would suggest to link Chl-a directly to blooms.

l 84: Maybe add Mattern et al (2010) to the list here, which compares several neighbourhood-based methods for model-satellite Chl-a comparison.

l 117: Maybe add that the model grid is used for this purpose here.

l 125: "chlorophyll": Shouldn't this also be Chl-a?

l 158: I would suggest to add one sentence about the difference between this product and the L4 product used in the comparison. Even just to mention that this is not the same dataset later used in the comparison.

Fig. 1: While "L4 observations" is no longer used in the text, it is still present in the figure and its caption. I am not against using the term but for consistency, the authors may want to change it.

Fig. 1: Negative exponents on the color bar are not fully in the superscript.

l 212: Use "Fig. 2a" instead of "a", I was at first confused what it was referring to.

l 216: "Only the pairing with the highest score is analysed further": In this particular example, or in general? What would happen if the 3-2 pairing scored 0.75 in Fig. 2b? I assume it would also become a match but the text suggests it would not.

Fig. 2: Nice example but it would be useful to more explicitly mention that 1 and 1 are matched. Currently only the bold text suggests it.

Fig. 2: The 1km distance bar should be moved closer to the objects in 2a.

l 236: In the answer to my previous comment, the authors mention that the observations are smoothed as well. This should be made explicit here.

l 278: When mentioning the threshold of 1000, it would also be good to mention the time resolution that is used here. Is a distance of one unit in time treated equally to one unit in space?

l 290: Maybe write "to _only_ capture events of interest" ("only" added by me) to emphasize why that would mean increasing the threshold.

l 293: "though higher values are present": In the data, in the L4 product, or both?

l 296: "In the paper": Here? I would suggest to rephrase to "here", "in this study" or "in this paper".

l 304: Why is a count of 30 an important limit?

l 304 "Furthermore, the smoothing applied needs to be reduced with increasing concentration thresholds because objects become smaller and are less frequent.": Here the logic appears to be backwards: smoothing happens before applying the threshold. So, increased smoothing decreases the concentration, requiring a lower threshold.

l 308: Remove "too".

l 308: "objects may be spurious": It is not really clear what this mean in this context. Wouldn't increasing the threshold remove objects that are less relevant?

l 309: "AMM7v11 analyses are on a ~7 km grid." This has been mentioned before.

Fig. 3, 4, 6, 7, 11: It would be useful to show grid lines in the panels

l 332: "centile": This is the short form and less commonly used word for "percentile". Especially since percentages are never used here, I would recommend to use the more general and likely better known term "quantile" instead.

l 332: "less than equal": Shouldn't this be "equal"?

l 333: "corresponding AMM7v11 centile": This needs to be described a bit better, is the "corresponding" centile/quantile the quantile of the same value or the quantile that is associated with the same chl-a concentration?

l 350: Maybe a better example (rather than a constant value) would be the presence of high chl-a features in the right place, i.e. "A lack of objects suggests the presence of a model bias but it does not provide any sense of whether the model is producing enhanced Chl-a (albeit below the chosen threshold) at the right location, or not."

l 360: "In this instance, though there is bias, it did not prevent the identification of objects in either fields to the extent where the results did not reflect the potential for the analyses to provide features which could be matched, paired and compared.": This sentence needs to be rephrased and simplified.

l 415: "From an interpretation perspective": It's not clear what this means, I would recommend not using ambiguous phrases like this.

Fig. 7: Because the description explicitly mentions the total number of objects and because the lines often overlap, it may make for a better plot to show the total number of objects and the number of matched objects, instead of matched and unmatched.

l 441: Where are the "South West Approaches"? I'd suggest to either describe the location in different terms or mark it on the map.

l 441: "There are areas, for example in the South West Approaches, where there appears to be a good level of consistency.": But this is for identified and not matched objects, correct? How about similar maps that show the probability of the grid cell being part of a match given that is is part of an identified object?

Fig. 10: Does the same color for AMM7v11 and L4 indicate that the objects matched?

l 566: Are these 26 days solely based on the first bloom object (l 501)?

l 587: "MODE results suggest that the AMM7v11 bloom objects are larger than those in the L4 product.": That appears a bit surprising given the results in Fig. 5 showing large L4 objects, the higher chl-a concentration in L4 (see distribution), and the result that AMM7v11 objects are numerous but small.

# References:

J. P. Mattern, K. Fennel, and M. Dowd (2010), Introduction and Assessment of Measures for Quantitative Model-Data Comparison Using Satellite Images, Remote Sensing. DOI: 10.3390/rs2030794.

Hide

RR by Anonymous Referee #2 (19 Apr 2021)

ED: Reconsider after major revisions (04 May 2021) by Andrew Moore

AR by Marion Mittermaier on behalf of the Authors (15 Jun 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (23 Jun 2021) by Andrew Moore

RR by Anonymous Referee #1 (30 Jun 2021)

Suggestions for revision or reasons for rejection

The most recent revisions have really improved the manuscript, made it easier to follow, and the updated Figures 6 and 11 are very helpful in illustrating the use of MODE and MDT in a real-world application. I only have one minor comment.

general comments

I like the addition of the "traditional" statistics but it would have been nice to incorporate them a bit better with the rest of the manuscript, that is, bringing them up when the MODE/MDT results are described and discussed.

specific comments

l 17: Add that the verification methods helped to reveal the following model results.

l 117: Maybe mention straight away that this model grid is coarser than the data grid and that the data is interpolated onto the model grid

l 219: "For the example blue object 1" -> "In the example in Figure 2(b), blue object 1"

Fig. 2: Should be "centroid distance" and not "centroid difference".

l 304: I would suggest to introduce the thresholds one sentence before they are used, i.e.: "three thresholds of 2.5, 4 and 6.3 mg m-3 were selected to be equally spaced in logarithmic space"

l 345: "but for higher thresholds (e.g. 6.3 mg m-3 )": Since only one is used here: "but for the higher 6.3 mg m-3 threshold"

l 352: "Statistics for matched pairs": I would remove the word "matched" here to not confuse the reader, as no matching in the MODE/MTD sense was performed. Or if it was, this needs to be elaborated on.

l 404: "(as seen in Figure 4 Figure 5)": Should just be Figure 4, I assume, or an "and" is missing.

l 409: "the quantile mapping functionality within MODE": Is there a reference you could add for a paper that describes this technique in more detail?

Fig. 6: This is a really nice plot showing the the results of MODE. I notice a few new objects in the western part of the domain, compared to the previous version of the manuscript, which I assume were dropped accidentally from the old version of the figure.

l 436: "suggest" -> "indicate"

l 530: Here and in the following line, the 25th percentile becomes the 0.25 quantile.

Fig. 11: Again, nice figure! Minor complaint: It is difficult to distinguish some of the colors, like the shades of purple, is there something that could be done to better match the temporal with the spatial location?

Hide

RR by Anonymous Referee #2 (04 Aug 2021)

Suggestions for revision or reasons for rejection

I thank the authors for revising the manuscript. The revised version reads improved. I can recommend the manuscript for publishing after small corrections (please see my specific comments).

Specific comments

L94-96 the sentence started with “A full traditional” could be slightly rephrased given the following sentence or removed.

L108-110: Currently sounds abrupt.
It would be nice to emphasize once again the aim of the study.

L298-299: please provide a reference on using log10 threshold, or please rephrase

L304: “were not log-transformed” instead of “were not transformed”

L306: could it be explained clearer?

L321: Should it be “pushing the concentration threshold too low” instead of “pushing the concentration threshold too high”? otherwise sounds contradictory.

L322: is something missing in the expression in the brackets?

L329-330: is there also a 4 km resolution L4 product?

L635: the following references on bloom phenology analysis in modelling could be provided: Ardyna et al. (2017), Hague & Vichi (2018), Pefanis (2021), Rohr et al. (2017), Song et al. (2010).

Typos/misprints

L73: citation format: remove the bracket before Shutler;

L84: please correct citation formatting: currently there are double brackets;

L209: mind the font used for “merged” and “one”;

L211: mind the font used for “matched”;

L404:”in Figure 4Figure 5” – please correct;

L529: mind the font used for “and”;

L566: should it be “in July” instead of “into July”?

References
Ardyna, M., H. Claustre, J.-B. Sallée, F. D’Ovidio, B. Gentili, G. van Dijken, F. D’Ortenzio, and K. R. Arrigo (2017), Delineating environmental control of phytoplankton biomass and phenology in the Southern Ocean, Geophys. Res. Lett., 44, 5016–5024, doi:10.1002/2016GL072428.

Hague, M., & Vichi, M. (2018). A link between CMIP5 phytoplankton phenology and sea ice in the
Atlantic Southern Ocean. Geophysical Research Letters, 45, 6566–6575. https://doi.org/10.1029/2018GL078061

Pefanis, V (2021), Loading of coloured dissolved organic matter in the Arctic Mediterranean Sea and its effects on the ocean heat budget. https://doi.org/10.26092/elib/646

Rohr, T., M. C. Long, M. T. Kavanaugh, K. Lindsay, and S. C. Doney (2017), Variability in the mechanisms controlling Southern Ocean phytoplankton bloom phenology in an ocean model and satellite observations, Global Biogeochem. Cycles, 31, 922–940, doi:10.1002/2016GB005615.

Song, H., Ji, R., Stock, C., Wang, Z (2010), Phenology of phytoplankton blooms in the Nova Scotian Shelf–Gulf of Maine region: remote sensing and modeling analysis, Journal of Plankton Research, Volume 32, Issue 11, 1485–1499, https://doi.org/10.1093/plankt/fbq086

Hide

ED: Publish subject to minor revisions (review by editor) (05 Aug 2021) by Andrew Moore

AR by Marion Mittermaier on behalf of the Authors (24 Aug 2021) Author's response Author's tracked changes Manuscript

ED: Publish as is (02 Sep 2021) by Andrew Moore

AR by Marion Mittermaier on behalf of the Authors (17 Sep 2021) Manuscript

Short summary

Regions of enhanced chlorophyll-a concentrations can be identified by applying a threshold to the concentration value to a forecast and observed field (or analysis). These regions can then be treated and analysed as features using diagnostic techniques to consider of the evolution of the chlorophyll-a blooms in space and time. This allows us to understand whether the biogeochemistry in the model has any skill in predicting these blooms, their location, intensity, onset, duration and demise.