On the efficiency of the hybrid and the exact second-order sampling formulations of the EnKF: a reality-inspired 3-D test case for estimating biodegradation rates of chlorinated hydrocarbons at the port of Rotterdam

Gharamti, Mohamad E.; Valstar, Johan; Janssen, Gijs; Marsman, Annemieke; Hoteit, Ibrahim

doi:https://doi.org/10.5194/hess-20-4561-2016

Articles | Volume 20, issue 11

https://doi.org/10.5194/hess-20-4561-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/hess-20-4561-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 20, issue 11

Research article

|

15 Nov 2016

Research article |

| 15 Nov 2016

On the efficiency of the hybrid and the exact second-order sampling formulations of the EnKF: a reality-inspired 3-D test case for estimating biodegradation rates of chlorinated hydrocarbons at the port of Rotterdam

Mohamad E. Gharamti, Johan Valstar, Gijs Janssen, Annemieke Marsman, and Ibrahim Hoteit

Download

Final revised paper (published on 15 Nov 2016)
Preprint (discussion started on 20 Apr 2016)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Nice theoretical study - disappointing application', Anonymous Referee #1, 11 May 2016
- AC1: 'Response to Reviewer 1 Comments', M.E. Gharamti, 20 Jul 2016
RC2: 'Interesting but limited in scope', Anonymous Referee #2, 20 May 2016
- AC2: 'Response to Reviewer 2 Comments', M.E. Gharamti, 20 Jul 2016
RC3: 'Comments on hess-2016-156', Anonymous Referee #3, 21 May 2016
- AC3: 'Response to Reviewer 3 Comments', M.E. Gharamti, 20 Jul 2016
AC4: 'Revised (Marked) Manuscript', M.E. Gharamti, 20 Jul 2016

Peer-review completion

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

ED: Reconsider after major revisions (30 Jul 2016) by Harrie-Jan Hendricks Franssen

AR by M.E. Gharamti on behalf of the Authors (01 Aug 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (02 Aug 2016) by Harrie-Jan Hendricks Franssen

RR by Anonymous Referee #2 (03 Sep 2016)

RR by Anonymous Referee #3 (09 Sep 2016)

Suggestions for revision or reasons for rejection

(1) I still have some concerns about the novelty of parts of the presented work. More than half of the results section deals with the comparison of EnKF and EnKF-OI and a sensitivity analysis on the weighting factors for EnKF-OI. This has already been presented in Gharamti et al. (2014) and I do not see why this discussion should be repeated in this manuscript (given that the conclusions and the overall model setup are very similar, see below). The authors should instead focus more on aspects that are built on top of the findings from Gharamti et al. (2014). For example, the authors mention that the sequential treatment of observations in the EnKF-OI scheme is a novelty in this work. However, if this is an innovative aspect of the paper why is this new method then not compared to the previous implementation? In the current state of the manuscript, the reader is left alone wondering if this change in the method really provides the supposed improvements.

(2) Additionally, a point that is stressed throughout the manuscript is that the authors intend to deal with a more realistic model setup. Real-world problem are generally characterized by uncertainty of subsurface parameters and boundary conditions, spatial heterogeneity, the presence of model structural errors and an increased level of complexity. However, most of the results presented in the paper are based on simplified model assumption, e.g., perfect knowledge on the flow field and hydraulic parameters and a lack of model dynamics and heterogeneity. Due to these perfect model assumptions, the usage of a 3D model (instead of a 2D case) is also not likely to add much complexity to the overall setup. The new results including uncertainty in hydraulic parameters (presented now in section 4.3) are quite interesting and relevant for the practical application of these kind of methods. So I would recommend the authors to extend these investigations and focus more on the practical issues of the method that arise under real-world conditions (as an ideal case was already discussed in their previous paper).

(3) Currently, section 4.3 also lacks some quantitative information that underpin the comparison between EnKF and EnKF-OI. It is mentioned, that EnKF-OI outperforms EnKF for a moderate uncertainty of hydraulic parameters but no numbers are given to judge the different performance. Additionally, results for EnKF are missing in Figure 18.

(4) Additionally, there is also a problem with the initialization of the assimilation experiments. In a real-world case, one would typically derive the initial concentration fields by simulating a warm up period with historical data until the assimilation of measurements starts at a certain time t0 (e.g., after a monitoring network was installed). However, what is done in this paper is that the initial concentration fields are derived by taking snapshots from a forward simulation for exactly the same time period that is used later in the assimilation experiments. The problem here is that the initial ensemble at time t0 already 'predicts' a considerable contamination along the whole (future) trajectory of the plume. On the contrary, the observations at time t0 state that the aquifer is completely clean initially. For a real-world situation this would be quite unrealistic because this huge discrepancy between observed and simulated concentration (clean versus totally contaminated) would heavily questions the credibility of the forward model, i.e. such a forward model would eventually be discarded due to the unrealistic bias. I would recommend the authors to account for this problem and to adapt their model initialization accordingly.

Hide

ED: Publish subject to minor revisions (Editor review) (17 Sep 2016) by Harrie-Jan Hendricks Franssen

AR by M.E. Gharamti on behalf of the Authors (26 Sep 2016) Author's response Manuscript

ED: Publish subject to technical corrections (15 Oct 2016) by Harrie-Jan Hendricks Franssen

Dear Dr Gharamti,

Thanks for the revised version of your manuscript "On the efficiency of the hybrid and the exact second-order sampling formulations of the EnKF: A reality-inspired 3D test case for estimating biodegradation rates of chlorinated hydrocarbons at the port of Rotterdam". I went through the revised manuscript and found that only technical corrections are needed which are listed below. I also think that the number of figures is very high and that it would be desirable to have not more than 15 figures. Please have a look if it is feasible to take out some figures which are less important.

Technical corrections:
- Often the plural “parameters” is used whereas it should be “parameter” in my opinion, like in “state-parameters estimation” which should be “state-parameter estimation”: L91, L136, L142, L152, L153 (2x), L180, L209, L213, L228, L482, L498, L506, L691

In other occasions the plural “states” should be used together with the plural “parameters”: L12, L176, L241, L555, L672, L684

In addition:
L53: “as they become available”
L54: skip “an”.
L145: The superscripts a and i were not specified.
L180: change to: “we will use a hybrid formulation (…)”
L194: change to: “in more detail”.
L324-L326: rephrase
L345: change to “reaction rates”
L373: change to “province”
L491: Information missing: for which alpha?
L516: Change to: “in the suggested performance”.
L525: space missing
L614: change to: “yields”
L616: change to “seems”
L618: change to: “appears”
L625: skip “a”

After these technical corrections the paper can be published. Congratulations!

Best regards,

Harrie-Jan Hendricks Franssen - editor -

Hide

AR by M.E. Gharamti on behalf of the Authors (16 Oct 2016) Author's response Manuscript

Short summary

The paper addresses the issue of sampling errors when using the ensemble Kalman filter, in particular its hybrid and second-order formulations. The presented work is aimed at estimating concentration and biodegradation rates of subsurface contaminants at the port of Rotterdam in the Netherlands. Overall, we found that accounting for both forecast and observation sampling errors in the joint data assimilation system helps recover more accurate state and parameter estimates.