Event selection and two-stage approach for calibrating models of green urban drainage systems

Broekhuizen, Ico; Leonhardt, Günther; Marsalek, Jiri; Viklander, Maria

doi:https://doi.org/10.5194/hess-24-869-2020

Articles | Volume 24, issue 2

https://doi.org/10.5194/hess-24-869-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-24-869-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 24, issue 2

Research article

|

26 Feb 2020

Research article |

| 26 Feb 2020

Event selection and two-stage approach for calibrating models of green urban drainage systems

Ico Broekhuizen, Günther Leonhardt, Jiri Marsalek, and Maria Viklander

Download

Final revised paper (published on 26 Feb 2020)
Preprint (discussion started on 08 Mar 2019)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Referee comment', Anonymous Referee #1, 07 Apr 2019
- AC1: 'Reply to referee #1', Ico Broekhuizen, 01 Jun 2019
RC2: 'Referee comment', Anonymous Referee #2, 18 Apr 2019
- AC2: 'Reply to referee #2', Ico Broekhuizen, 01 Jun 2019

Peer-review completion

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

ED: Reconsider after major revisions (further review by editor and referees) (03 Jun 2019) by Roberto Greco

AR by Ico Broekhuizen on behalf of the Authors (15 Jul 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (24 Jul 2019) by Roberto Greco

RR by Anonymous Referee #2 (21 Aug 2019)

Suggestions for revision or reasons for rejection

The authors have provided a thorough response to the first round review comments, and at parts revised the manuscript accordingly. Leaving out some of the content present in the original manuscript has improved the clarity of the revised manuscript. I my view there would be still be room for a more concise presentation to highlight the main findings of the research.

Documentation of the two-stage calibration (impermeable vs. permeable parameters) deserves to be published to a wider audience, currently the results are in my understanding only available in the licentiate thesis of the first author and in a brief form in a conference extended abstract.

The authors have retained the analysis on +/- 40% perturbed discharge data and the level of model discretization. In my view, as already expressed in my first review, the results related to these analyses do not have much novelty value and are partly trivial. The authors argue the novelty by claiming that earlier studies have predominantly addressed impervious catchments instead of ‘greener’ urban catchments. I am not totally convinced this is the case. In page 2 lines 20-21 the authors refer to three earlier studies with regard to spatial model resolution assessments, and state below that ‘these investigations used predominantly impervious catchments’. Strictly speaking, this is true, but I find it a bit misleading as out of six catchments addressed in the referenced studies two represent the same level of imperviousness as the study area in the current manuscript. Catchment 3 of Krebs et al., 2014, imperviousness 19 % (p. 483, Section 2 Study site and data) and Sucy catchment of Petrucci and Bonhomme, 2014, imperviousness ~20% (p. 1005, Figure 6b). Also, the authors point out that the variability of the calibrated rainfall multiplier values is larger for low-resolution than for high resolution models (p. 17, lines 1-2) and attribute this to ‘rainfall multipliers appearing to behave in a more physical way’ (p. 17, line 9). This sounds like an obvious result as the low resolution models, unlike the high resolution models, have another parameter directly affecting the volume of runoff (‘Percentage runoff routed from impervious to pervious’, Table 3, p. 10). In page 14 lines 6-8 the authors state themselves that while in Dotto et al. (2014) the mismatch between rainfall and discharge volumes was controlled by the calibrated value of percentage imperviousness in the current paper the same impact is obtained by manipulating rainfall multiplier values. Now in low-resolution models of the current manuscript, in essence, both of these controls are present (rainfall multipliers and ‘Percentage runoff routed from impervious to pervious’), which leads to a larger variability in any of these two controls individually, which is not surprising.

I would like to see this manuscript published but I feel a bit frustrated to see the most valuable part of the research results being mixed with the rather inconclusive results related to other sources of uncertainty. The latter are often trivial or confirm earlier findings of other research. In my mind, the manuscript would certainly benefit from a tighter framing and focus. In its current state most of the Conclusions address the results related to performance differences between one-stage and two-stage calibrations, or alternatively the relatively obvious results about the relationships between rainfall multipliers and the scaling coefficient of flow series (or variation of rainfall multipliers between high- and low-resolution models). Hardly any conclusions are drawn on the selection of calibration events (P_sum, PI_mean, PI_30m etc.), which according to the title of the manuscript should be in the focus. In my view this should be fixed by changing the title to emphasize the one-stage / two-stage calibrations and revising the contents accordingly. Should the authors wish to keep the original title, then the main results and conclusions should be aligned with it.

Specific (technical) comments

p. 6, line 4: Correct the reference Fuentes-Andino…

p.7, line 18: “..values of certain event characteristic…”, vague language, replace with e.g. ‘given event characteristic’

p. 8, line 4: Not totally clear what ‘Doing this’ refers to. Consider rephrasing the sentence.

p. 11, line 23: Word ‘Figure’ missing from Figure 2.

p. 17, line 18: “When switching from the high resolution to the low-resolution model the single-stage CSs were no longer able to predict up to 5 events” Consider rephrasing, hard to follow a sentence.

p. 19, lines 10-11: “inter-CS variation for the same events varies from 0.15 to 1.25.” The difference of 1.25 is hard to see from Figure 6, as the y-scale is cut at zero.

p. 21, line 7: “more less consistent than..” should probably read ‘less consistent result’

p. 22, Table 8: not obvious what all column titles mean, what is ‘Clip mean NSE’?

Hide

RR by Anonymous Referee #1 (27 Aug 2019)

ED: Publish subject to revisions (further review by editor and referees) (04 Sep 2019) by Roberto Greco

AR by Ico Broekhuizen on behalf of the Authors (16 Oct 2019) Author's response Manuscript

ED: Publish subject to revisions (further review by editor and referees) (29 Oct 2019) by Roberto Greco

ED: Referee Nomination & Report Request started (28 Nov 2019) by Roberto Greco

RR by Anonymous Referee #2 (27 Dec 2019)

Suggestions for revision or reasons for rejection

Dear Authors,

I am generally happy with the revised manuscript, in my opinion it offers now a much crisper presentation of the most valuable results of this research. I recommend acceptance of the manuscript with some minor changes/clarifications.

Title: Consider rephrasing the title, ‘calibration of green urban drainage models’ sounds a bit strange. What is a ‘green model’?

P. 3, lines 23-24: “This two-stage calibration has not been investigated for urban drainage models,…” Refer to the licentiate thesis / conference abstract where this idea was introduced.

P. 3, line 27: “…of the paper that follows is…” Perhaps just 'this paper' instead of 'the paper that follows'?

Table 1 (and elsewhere): Should QV rather be runoff depth as its unit is mm?

P. 8, lines 9-12: Remove parentheses. Entire sentences should not appear enclosed in parentheses in the middle of the text. Rephrase paragraph if needed.

P. 10, lines 24-26: Do the cited articles present results on one-stage vs. two-stage calibrations? I presume not. Rephrase the text so that it does not sound like the cited research would confirm results on one-stage/two-stage calibrations.

P. 11, last line: Remove parentheses. Rephrase if needed.

P. 13, line 2: “..from 0.48 to 2.92,…” Consider adding a couple of sentences to further elaborate this. I do not think the options you offer are necessarily equally likely. This probably is not a result of a (systematic) flow gauging error as it varies from an event to another. It well could be, as you suggest, attributed to differences between the gauge rainfall and the catchment mean rainfall. Alternatively it could result e.g. from the model’s deficiency in describing correctly losses/stored water for different events.

P. 13, lines 7-8: “Secondly, decreasing or increasing all flow rates by 40% prior to calibration changed the average rainfall multipliers by -37% and +33% respectively.” I still think this is rather obvious and should be omitted.

P. 15, lines 3-4: “…events, but produced wrong timing, which was also reflected in the peak flow and volume errors.” Not sure I understand this. Form the way PFR was defined (p. 9, lines 21-23) it seems that timing should not be an issue as the event peak flow value is picked at the time it occurred and hence a timing difference between modelled and observed flow should not be important? The same applies for flow volume – timing is not an issue as long as the time window encloses the entire event for both modelled and observed values?

P. 15, lines 20-21: “…for most events flows were underestimated, which may be (at least partially) attributed to the discrepancies between observed rainfall and flow found in the calibration phase (see Sect. 3.2.2).” Not sure I understand this either. The idea of introducing rainfall multipliers was to fix the potential mismatch between gauge rainfall and catchment mean rainfall. Now why would the discrepancy between gauge rainfall and catchment rainfall result in consistently underestimating (and not overestimating) the flow? Please clarify.

P. 15, line 24: “inter-CS variation for the same events varies from 0.15 to 1.25.” What are these values? Not quite clear to me.

P. 15, lines 31-33: “…model, while from the two-stage CSs only T32S_D_prec satisfactorily predicted fewer events, and T32S_P_sum, T32S_Q_max, and T32S_QV_ppP actually predicted a higher number of events satisfactorily.” This probably refers to comparison between HR and LR models, but in my mind it is not quite clear. Consider rephrasing.

P. 16, line 3: “…with this metric: First,…” replace the colon with a full stop.

P. 18, line 4: “…, the NSE…” Should this read ‘the event mean NSE’, i.e. does it refer to the mean NSE value over all calibration events (instead of NSE values over individual events)? If you change the wording, check also elsewhere (e.g. 3.5, Conclusions) for consistency.

P. 19, lines 7-8. “ By contrast, in the validation phase the NSE was better for the single-stage CSs.” This refers to the event mean NSE? But not to the ‘Joint NSE’ (Table 6)? Should this be somewhow discussed? At times it is difficult for the reader to follow which performance criterion is referred to, in particular as they can give mixed messages.

P. 20, line 4: “NSE values” Does this mean ‘event mean NSE values’?

Hide

ED: Publish subject to minor revisions (review by editor) (07 Jan 2020) by Roberto Greco

AR by Ico Broekhuizen on behalf of the Authors (17 Jan 2020) Author's response Manuscript

ED: Publish as is (24 Jan 2020) by Roberto Greco

Short summary

Urban drainage models are usually calibrated using a few events so that they accurately represent a real-world site. This paper compares 14 single- and two-stage strategies for selecting these events and found significant variation between them in terms of model performance and the obtained values of model parameters. Calibrating parameters for green and impermeable areas in two separate stages improved model performance in the validation period while making calibration easier and faster.