Comment on “Can assimilation of crowdsourced data in hydrological modelling improve flood prediction?” by Mazzoleni et al. (2017)

Viero, Daniele P.

doi:https://doi.org/10.5194/hess-22-171-2018

Articles | Volume 22, issue 1

https://doi.org/10.5194/hess-22-171-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/hess-22-171-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 22, issue 1

Comment/reply

|

10 Jan 2018

Comment/reply |

| 10 Jan 2018

Comment on “Can assimilation of crowdsourced data in hydrological modelling improve flood prediction?” by Mazzoleni et al. (2017)

Daniele P. Viero

Download

Final revised paper (published on 10 Jan 2018)
Preprint (discussion started on 01 Mar 2017)
Corresponding article

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review of the Comment on "Can assimilation of crowdsourced data in hydrological modelling improve flood prediction?'' by Mazzoleni et al. (2017)', Maurizio Mazzoleni, 27 Mar 2017
RC2: 'Review of Manuscript Number: hess-2017-102', Anonymous Referee #2, 25 Apr 2017
AC1: 'Author respose to reviewers' comments', Daniele P. Viero, 20 May 2017

Peer-review completion

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

ED: Reconsider after major revisions (further review by Editor and Referees) (02 Jun 2017) by Stacey Archfield

AR by Daniele P. Viero on behalf of the Authors (10 Jul 2017) Author's response Manuscript

ED: Publish subject to revisions (further review by Editor and Referees) (28 Jul 2017) by Stacey Archfield

ED: Referee Nomination & Report Request started (03 Aug 2017) by Stacey Archfield

RR by Anonymous Referee #3 (12 Aug 2017)

Suggestions for revision or reasons for rejection

Thanks for the opportunity to review the revised version o this comment. I read both this version and the preceding discussion with interest. Honestly, I have somewhat split feelings about this comment. On the one hand some valuable points are made, but on the other hand the 'comment-aspects' are not fully clear.

As I see it the author makes two main comments: 1) the use of synthetic data for the crowdsourced observations and 2) the model choice and application. While the author is very critical about the first point, I would disagree to phrase this as general as done in the comment. Studies using synthetic data can actually be quite informative to investigate the question on how valuable such data potentially could be if they would be available. I would argue this is a suitable approach can actually provide guidance on how to collect crowed sourced data (see also van Meerveled et al, in review, HESS-D, as an example of this approach). The author needs to provide more convincing arguments why the approach in general is not suitable or on where exactly he sees short-comings of the particular implementation of this approach in the study by Mazzoleni et al.

The second point I actually find more interesting. The author nicely provides reasons on why a so called physical model is not as physical as one might think. I find this discussion really helpful, although it could be a bit more to the point. I can clearly sympathize with the argument that the use of such a 'physical but still not so physical' model has implementations for the result in a study which heavily relies on model calibration, and probably the use of a simpler model by Mazzoleni et al. could have been appropriate. However, it is not clear from the comment in which way the author thinks this could have influenced the results.

From the comment it is clear that certain aspects of Mazzoleni et al. could have done differently or have been described clearer. However, to be really useful a comment has to be specific and raise issues of general importance. As argued above, I do not agree with the fundamental critic against the use of synthetic data. While point 2 still could be of general interest, it is not yet formulated in such a way. In the current form it mainly describes the details but misses to frame this i a more general discussion on which model to use when, including the consequences of using a too complex model.

T summarize, while the comment raises an important point which could be of general interest, some significant work is needed to make the comment as useful as it could be.

Reference:

van Meerveld, I., Vis, M., and Seibert, J.: Information content of stream level class data for hydrological model calibration, Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2017-72, in review, 2017.

Hide

RR by Maurizio Mazzoleni (01 Sep 2017)

Suggestions for revision or reasons for rejection

The author clearly improved the manuscript and I really appreciate his effort. The structure of the comment is now clear and the main objective well described. I do believe that author critical remarks will be helpful for future publications in the same area. However, I still have a couple of comments/suggestions regarding section 3 (use of CSD in operation flood forecasting). I think this section needs more clarifications. In particular, from the abstract and introduction (page 2 lines 1-3) it seems that the main objective of the comment is to discuss the effects of assimilation of unreliable synthetic CSD derived from a poorly calibrated model on the assimilation performances. Nonetheless, section 3 provides a qualitative analysis on the effect of model structural uncertainty in data assimilation, which is different than the original scope of this comment. The author need to clarify this issue or to better frame the scope of the comment. In addition:

• Page 5, lines 25-30: Is the author referring to the assimilation of synthetic or real CSD? I do agree that synthetic CSD estimated at different points than the calibration ones may not be accurate when model is poorly calibrated. This concept was already reported in section 2.3. On the other hand, I do not see the point of this paragraph if the author is referring to real CSD. Obviously, a traditional physical sensor will provide more reliable observations than CSD if located at the same point. However, the benefit of CSD is in their spatial distribution and availability in points where physical sensors are not available, as stated already in the original paper. I suggest the author to clarify which type of CSD (synthetic or real) are considered in section 3.
• Page 6, lines 18-24: It is stated in the comment that data assimilation can be used only at calibration points in case of poorly calibrated model. I have some doubts about this statement (in particular page 6 lines 20-22). It is worth noting that even in case of not-properly calibrated model (which is the case of semi-distributed and distributed models) assimilation of reliable observations can help improving model performances. The author should include that a proper estimation of the model error, expressed by means of the error covariance matrix in linear DA or model spread in Ensemble method, is necessary to ensure an appropriate assimilation process in case of poor model calibration. There are many studies in which observations are assimilated at internal points different than the calibration points. This is the case of assimilation of remote sensing observations in case of distributed model calibrated only in few discharge locations (when available). I suggest the author to improve this part.
• Page 7, lines 28-30: I do not agree with this statement. It is because of the random behaviour and involvement of citizens that CSD location cannot be determined a-priori. CSD can be provided at any point of the basin/river and not only at calibration ones. There has been many studies in which mobile applications are developed to correctly estimate river velocity or flow (e.g. Lüthi et al. 2014) by using river cross section (which can be assumed rectangular). Definitely, such tools will provide an uncertainty estimation of the flow characteristics of the river which may lead to a wring update of the model states at interior points. The hope is that more reliable tools for accurately measuring river flow will be developed in the next years.

I have a final question to the author. Since CSD can be assimilated only at points when model is calibrate and physical sensors (more reliable than CSD) are already installed at these points, is the author implicitly suggesting that CSD from citizens should not be used for improving flood predictions even in case that system states are well represented?

I hope these suggestions will help the author to further improve his valuable contribute to the use of CSD from citizens for improving flood prediction.

Reference:
Lüthi, B., Philippe, T. and Peña-Haro, S.: Mobile device app for small open-channel flow measurement, International Environmental Modelling and Software Society (iEMSs) 7th Intl. Congress on Env. Modelling and Software, San Diego, CA, USA, 2014.

Hide

ED: Publish subject to revisions (further review by Editor and Referees) (18 Sep 2017) by Stacey Archfield

The manuscript has now undergone 2 rounds of review. The first set of review comments were considered major and, therefore, the revised manuscript was sent out for review again after the author had made substantial changes to the manuscript.

In this round of reviews, Reviewer #2 notes that there has been significant improvement to the revised manuscript and recommends that the manuscript be accepted with minor revision. Reviewer #1 notes there are interesting points raised in this comment but felt that there was further major revision (and review) needed for the manuscript to be accepted.

In looking at both reviewers comments and my own review of the revised manuscript, the main points of this commentary still need clarification and be generalized enough to make the comment of interest to a wide audience. Reviewer #2 notes that this quite possible but that major revision is still needed.

To help the author clarify the contribution, the reviewers have noted their interpretations of the major points of this commentary and provide specific comments about how this can be done. I offer the same here to help with revisions: In my understanding of the contribution of this comment, I find of most interest the point that models are often called upon for applications that require information to make decisions of high societal value - such as flood forecasting - and that we have a responsibility as hydrologists to understand the limits and applicability of our models and the data that drive those models. As crowd-sourcing data (CSD) can now be used as a further tool to enhance hydrologic modeling efforts, out community needs to consider the use of CSD in our responsible assessment of the applicability of models to answer questions of such importance and relevance. I see the author using a recent paper on CSD, study area, and model to make this comment.

I believe that an additional revision effort, will continue to improve the general interest of the paper and elevate the impact of the comment. For this reason, I am recommending additional revision before the manuscript can be accepted.

I thank again the author and reviewers for the productive discussion and progress towards publication.

Hide

AR by Daniele P. Viero on behalf of the Authors (27 Sep 2017) Author's response Manuscript

ED: Referee Nomination & Report Request started (23 Oct 2017) by Stacey Archfield

RR by Anonymous Referee #3 (06 Nov 2017)

ED: Publish subject to minor revisions (review by editor) (21 Nov 2017) by Stacey Archfield

The reviewer of the revised comment has determined that the manuscript is a worthwhile contribution to the literature and is ready for publication. I have now read the revised comment, and while I agree, the comment is a worthwhile contribution, there are still areas of the comment that require editorial revisions before the manuscript can be accepted.

I now ask the author to address these comments. I will assess that the comments have been addressed before final acceptance of the manuscript for publication.

- Please proofread thoroughly the entire manuscript.

- Abstract: The abstract does not mention flood forecasting nor does it seem to capture the main points of the comment. I would look to the summary of the comment, particularly the last 2 paragraphs, which contains some nice text about the main points of the comment.

- p. 1, line 3: Change “moment” to “time” and delete the word “actually”

- p. 1, line 10: You do not use the word “unreliable” later to describe the use of synthestic CSD. I would remove this sentence here and replace with some of the points made in the summary.

- p. 1, lines 15-16: Move the phrase “in order to produce the most accurate flood predictions” to the start of the start of the sentence.

- p. 1, line 18: Delete “, and to explore different kind of” and “as well”

- Introduction, paragraphs 1 and 2: Abbreviate CSD after the first use of crowdsourced data (line 19) and then use CSD after that (line 21).

- p.1, line 20: Change “unexpected” to “exciting”

- p. 2, line 1: Move the reference to Mazzoleni to after the phrase “In pioneering applications”

- p. 2, line 3: Change to read “In this recent work”

- p. 4, line 27: Change “hard” to “difficult”

- p. 5, line 14: What is meant by “closing section” here? Rephrase using a hydrology term.

- p. 5, line 18, Change to read “In their…”

- p. 5, line 20: Provide the reason for why this was not surprising. This will help reiterate the point you are making in this paragraph.

- p. 5, line 22: Change to read “…the model must well represent the physics…”

- p. 5, line 24: I do not see where OSSE has been spelled out before it is abbreviated here.

- p. 5, lines 25-29: These sentences are still a bit unclear. This text should be more clear in describing one of the central points of your comment. Provide more detail so that the reader is able to easily grasp the point.

- p. 5, line 33: Delete the last sentence of this paragraph.

- Section 3 is nicely written.

- p. 6, line 3: Change to read: “However, this comment points out that attention must be paid…”

- p. 6, line 4: Please add a reference to some literature that use ensemble based methods for flood forecasting.

- p. 8, line 5: Change to CSD where “crowdsourced data” is used.

- Wording needs to be softened in several areas of the comment:

p. 1, line 21: Delete “severe”
p. 1, line 1: Delete “in fact” and expand on why a lack long time series of CSD data are a further limitation.
p. 2, line 16: Change to read “…, which affects the interpretation of the usefulness of CSD.”
p. 5, line 21: Change “claimed” to “argued”
p. 5, line 24: Delete “As a matter of fact…”
p. 5, line 30: Change to read “It is possible that such an inconsistency….”
p. 5, line 31-32: Change to read “Therefore, additional care must be taken…”
p. 7, line 6: Change to read “As an alternative for…”
p. 7, line 27-28: Change to read “and flow routing processes face limitations…”

Hide

AR by Daniele P. Viero on behalf of the Authors (23 Nov 2017) Author's response Manuscript

ED: Publish as is (26 Nov 2017) by Stacey Archfield

AR by Daniele P. Viero on behalf of the Authors (27 Nov 2017)