A parallel workflow implementation for PEST version 13.6 in high-performance computing for WRF-Hydro version 5.0: a case study over the midwestern United States

Wang, Jiali; Wang, Cheng; Rao, Vishwas; Orr, Andrew; Yan, Eugene; Kotamarthi, Rao

doi:https://doi.org/10.5194/gmd-12-3523-2019

Articles | Volume 12, issue 8

https://doi.org/10.5194/gmd-12-3523-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-12-3523-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 12, issue 8

Development and technical paper

|

13 Aug 2019

Development and technical paper |

| 13 Aug 2019

A parallel workflow implementation for PEST version 13.6 in high-performance computing for WRF-Hydro version 5.0: a case study over the midwestern United States

Jiali Wang, Cheng Wang, Vishwas Rao, Andrew Orr, Eugene Yan, and Rao Kotamarthi

Download

Final revised paper (published on 13 Aug 2019)
Supplement to the final revised paper
Preprint (discussion started on 29 Nov 2018)
Supplement to the preprint

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Referee Comment', John Doherty, 06 Dec 2018
RC2: 'Referee comment', Anonymous Referee #2, 16 Jan 2019
AC1: 'Response to reviewers' comments and revised manuscript with tracked changes', Jiali Wang, 09 Mar 2019

Peer-review completion

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

AR by Jiali Wang on behalf of the Authors (09 Mar 2019) Manuscript

ED: Referee Nomination & Report Request started (24 Mar 2019) by Wolfgang Kurtz

RR by Anonymous Referee #2 (08 Apr 2019)

Suggestions for revision or reasons for rejection

I acknowledge that the authors made a big effort to improve the manuscript and agree it is substantially improved. Nevertheless, I still have some concerns.
Concerning the analysis of computational benefits of parallel PEST on HPCs (new Section 4.2), the analysis performed is interesting, but to me it should be still improved, because does not yet respond clearly to the main question: “How can I use the nodes I have available in the most efficient way?” If I understand this section, based on Table 3 the Test 5 is better than the Test 1 because: Test 1) 23 workers x 2 nodes = 46 nodes and 103 min; Test 5) 6 workers x 6 nodes = 36 nodes and only 86 min. On the one hand, this result should be highlighted. But, on the other hand, the same result cannot be separated from considering the extent of the WRF-Hydro domain (from this point of view, I note that information about computational domain is still missing). Therefore, in general I would say that: first, a scalability analysis of WRF-Hydro over the specific computational domain should be performed; then, this scalability analysis should be used as a preliminary, but essential piece of information to provide comprehensive indications about the “optimal” configuration of the system for the analysed case study, given the threshold of the computational resources available.
Concerning my doubts about the calibration procedure, the explanation provided: “The reason is that the water contributions for these stations are from a larger river basin (Mississippi River) than we included in our current study area” confirms that the calibration performed over stations from 2 to 4 does not make too much sense. Of course enlarging the study area to include the MRB will improve results, and this must be done if the authors still want to consider stations from 2 to 4. Another option, maybe more feasible for the authors, is to skip stations from 2 to 4 and consider stations from 5 to 8 not only for “transferability”, but also for validation (but this must be done in a persuasive way). However, since from the first version of the manuscript it was not clear that the study area “covers only half of the MRB”, and in the second version the authors only hint at it, it is very useful, for the sake of clarity, that the characteristics of all the catchments upstream the considered stations are clearly showed in the paper, both in a Table (e.g., indicating extent of the catchment and other main features) and in Figure 1, highlighting the borders of the catchments.
I hope my comments can help further improving the manuscript.

Hide

ED: Reconsider after major revisions (21 Apr 2019) by Wolfgang Kurtz

AR by Jiali Wang on behalf of the Authors (16 Jun 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (19 Jun 2019) by Wolfgang Kurtz

RR by Anonymous Referee #2 (03 Jul 2019)

ED: Publish subject to technical corrections (12 Jul 2019) by Wolfgang Kurtz

AR by Jiali Wang on behalf of the Authors (17 Jul 2019) Author's response Manuscript

Short summary

WRF-Hydro needs to be calibrated to optimize its output with respect to observations. However, when applied to a relatively large domain, both WRF-Hydro simulations and calibrations require intensive computing resources and are best performed in parallel. This study ported an independent calibration tool (parameter estimation tool – PEST) to high-performance computing clusters and adapted it to work with WRF-Hydro. The results show significant speedup for model calibration.