Improvement of hydrological model calibration by selecting multiple parameter ranges

Wu, Qiaofeng; Liu, Shuguang; Cai, Yi; Li, Xinjian; Jiang, Yangming

doi:https://doi.org/10.5194/hess-21-393-2017

Articles | Volume 21, issue 1

https://doi.org/10.5194/hess-21-393-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Special issue:

Modeling hydrological processes and changes

https://doi.org/10.5194/hess-21-393-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 21, issue 1

Research article

|

24 Jan 2017

Research article |

| 24 Jan 2017

Improvement of hydrological model calibration by selecting multiple parameter ranges

Qiaofeng Wu, Shuguang Liu, Yi Cai, Xinjian Li, and Yangming Jiang

Download

Final revised paper (published on 24 Jan 2017)
Preprint (discussion started on 20 Jun 2016)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Comments on the manuscript hess-2016-275', Anonymous Referee #1, 04 Aug 2016
- AC1: 'Response to the comments of Referee #1', YI CAI, 26 Aug 2016
RC2: 'Comment on "Improvement of hydrological model calibration by selecting multiple parameter ranges"', Anonymous Referee #2, 05 Aug 2016
- AC2: 'Response to the comments of referee #2', YI CAI, 28 Aug 2016

Peer-review completion

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

ED: Reconsider after major revisions (29 Aug 2016) by Giuliano Di Baldassarre

AR by YI CAI on behalf of the Authors (06 Sep 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (21 Sep 2016) by Giuliano Di Baldassarre

RR by Anonymous Referee #2 (22 Oct 2016)

Suggestions for revision or reasons for rejection

Comment on the revised version: “Improvement of hydrological model calibration by selecting multiple parameter ranges”

Authors have considered all the comments risen in the previous review and they have replied in a satisfying way to almost all of them. The quality of the overall manuscript has been improved relative to the original submission. However I think a further check to the text will further improve the fluency. In my opinion the manuscript can be considered for publication after some minor revision.
Abstract is not well structured and not clear. Hereafter some comments.
The sentence at line 16 is not easy to read. I suggest to limit it as “…initial parameter range and distribution type (i.e. normal, exponential and uniform distributions).” In the following sentences Authors refer to specific indexes, such as ENS, MINR, MAXR that are not know by the readers. They cannot guess their meaning. This happens again at line 25 with Im. Please, revise the overall abstract in order to summarize the concepts of your work, the elements of novelty but without referring to specific parameters or indexes that are defined only in the main text.
P3,L0: special? I think “specific” here is more appropriate.
P10, L24: I am still not convinced that a variation in term of model efficiency of nearly 0.01 could be defined as “much greater”. I do agree with the importance of reducing the NSE variability, but I suggest to reduce the emphasis on this point.
P10, L33: “maximum and minimum ENS values…” and “ using the proposed” (please check for this in many cases along the manuscript); please remove “respectively”.
Figure 12: I appreciate the new figure but I wonder why there is such a brutal change on the min values after 100 runs. It’s a bit strange that GA and the proposed method produce essentially the same values till 100 run and then they suddenly diverge. Do the Authors have any explanation for that? Can it be related to the sampling number of parameter values?
What about the computational time required by the two methodologies? Are there significant differences?
Conclusion: L10, please remove “for karst area”, it’s specify at the previous line.

Hide

RR by Anonymous Referee #1 (30 Nov 2016)

Suggestions for revision or reasons for rejection

Overview
--------
This study touches upon the problematic of obtaining inadequate calibration results in hydrological modelling as a result of inappropriate parameter ranges. To address this issue the authors present a new approach to select parameter ranges based on probability distribution characteristics of the parameter space. The methodology is divided in three main steps: (i) determination of the probability distribution of the calibrated parameter values, (ii) adjustment of the initial parameter ranges of each parameter individually by finding the ranges of maximum and minimum probability density for a given cumulative frequency, and (iii) resolving the impact of the adjusted parameter range on other parameters and keeping the modified range only if it has a positive impact on the calibration of the other parameters.

In order to demonstrate the application of the proposed methodology the authors calibrated the Xinanjiang model for the Chaotianhe River catchment. Out of 10 model parameters that needed to be calibrated, the authors found that 3 parameters follow a normal distribution and the other 7 an exponential distribution. They then proceed to constrain the parameter ranges following the proposed approaches for each of the distributions and test a number of cases involving different combinations of parameters kept at the initial ranges and having optimized ranges. Finally, the authors conclude that the proposed approach make model calibration more effective.

General comments
----------------
This revised (Revision 1) manuscript presents a new methodology to determine adequate parameter ranges in order to improve calibration results in hydrological modelling. The topic is suitable and interesting for the Hydrology and Earth System Sciences readership.

In the revision process, the authors clarified a number of my earlier comments, most notably on the selection of probability distribution functions, benefits of the proposed method, and required sample size. However, some aspects of the manuscript need further work. Please find my more detailed comments about the revised version of the manuscript. I recommend that the manuscript is accepted after revisions based on the comments provided below. Note that the lines referred below correspond to the document "Revised manuscript with no changes marked".

Main comments
-------------
1) Even if the language was generally improved in this revised version some sections still require further work. More specifically, the Introduction section needs most work. This is not a trivial issue as it affects the readability of the manuscript and can discourage potential readers. I would recommend the authors to have a native English speaker proof-check the manuscript.

2) In the revision the authors clarified the magnitude of the efficiency improvement produced by the proposed method. In their response to my previous comment the authors recognize that the modes performance increase might be due to use of the Xinanjiang model for the case study. As I see it a case study should be selected to demonstrate the benefits of the proposed methodology. Following this reasoning I wonder the rationale behind selecting the specific catchment and hydro-meteorological model for this study. Please clarify.

3) Page 4, line 8. The authors report that areal precipitation was calculated bason on the Thiessen polygon method. This method is known to introduce significant errors in catchments with complex topography. Since the study area appears to be quite mountainous I wonder about the rationale behind using the Thiesse polygon method. Please clarify.

Specific comments
-----------------
1) Page 4, line 10: The most common and accepted term is "hydro-meteorological".

Hide

ED: Publish subject to minor revisions (further review by Editor) (01 Dec 2016) by Giuliano Di Baldassarre

AR by YI CAI on behalf of the Authors (10 Dec 2016) Author's response Manuscript

ED: Publish as is (14 Dec 2016) by Giuliano Di Baldassarre

AR by YI CAI on behalf of the Authors (15 Dec 2016)

Short summary

We proposed a method to calibrate hydrological models by selecting parameter range. The results show the probability distribution can be used to determine the optimal range of a single parameter. Analysis of parameter sensitivity and correlation is helpful to obtain the optimal combination of multi-parameter ranges which contributes to a higher and more concentrated value of the objective function. The findings can provide references for enhancing the precision of hydrological process modelling.