Reduction of predictive uncertainty in estimating irrigation water requirement through multi-model ensembles and ensemble averaging

Multsch, S.; Exbrayat, J.-F.; Kirby, M.; Viney, N. R.; Frede, H.-G.; Breuer, L.

doi:https://doi.org/10.5194/gmd-8-1233-2015

Articles | Volume 8, issue 4

https://doi.org/10.5194/gmd-8-1233-2015

© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/gmd-8-1233-2015

© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 8, issue 4

Methods for assessment of models

|

29 Apr 2015

Methods for assessment of models |

| 29 Apr 2015

Reduction of predictive uncertainty in estimating irrigation water requirement through multi-model ensembles and ensemble averaging

S. Multsch, J.-F. Exbrayat, M. Kirby, N. R. Viney, H.-G. Frede, and L. Breuer

Download

Final revised paper (published on 29 Apr 2015)
Preprint (discussion started on 10 Nov 2014)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC C2669: 'Referee comment MS 7525-2014, Multsch et al.', Anonymous Referee #1, 15 Dec 2014
- AC C2704: 'Response to anonymous Referee #1 (gmdd-7-C2669-2014)', Sebastian Multsch, 19 Dec 2014
  - AC C3249: 'Response to anonymous referee #1, corrected version', Sebastian Multsch, 04 Feb 2015
RC C2866: 'gmd-2014-189', Anonymous Referee #2, 05 Jan 2015
- AC C2976: 'Reply to “Geosci. Model Dev. Discuss., 7, C2866–C2868, 2015”', Sebastian Multsch, 19 Jan 2015

Peer-review completion

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

AR by Sebastian Multsch on behalf of the Authors (17 Feb 2015) Author's response Manuscript

ED: Referee Nomination & Report Request started (26 Feb 2015) by James Annan

RR by Anonymous Referee #2 (11 Mar 2015)

RR by Anonymous Referee #1 (14 Mar 2015)

Suggestions for revision or reasons for rejection

The authors implemented several changes to the manuscript but I’m still not convinced that the method of using „Reliability Ensemble Averaging“ (REA) represents in this specific case any advantage to established methods of model calibration and validation. REA requires an estimate of the reliability of the different methods used in the ensemble to define the weight of the different ensemble members for the ensemble mean. The authors use adjusted class-A pan evaporation measurements fort his purpose. I still cannot see the value of using an ensemble mean where results obtained with relatively accurate methods like Penman-Monteith are merged with results from PET calculation methods that are known to be less precise. The common procedure would certainly be to split the class-A pan measurements into two parts, using the first set of measurements to identify the method which performs best and to calibrate this method to the class-A pan measurements. Then, the second set of measurements can be used for validation. This procedure provides an estimate of the accuracy at the locations where class-A pan measurements were taken and for the whole study area as well, assumed that the locations are representative for the whole study area.

The approach used by the authors requires more effort because each simulation has to be repeated several times with other PET calculation methods but I cannot see any advantage of the REA ensemble mean to an estimate which is based on the best performing method (likely Penman-Monteith) calibrated to observations. Furthermore, the main conclusion of the study, that uncertainty caused by model structure (different ET calculation methods) is higher than uncertainty caused by different sets of crop coefficients, depends certainly on the selection of ET methods considered in the study and on the selection of the study region. Therefore, this result cannot be generalized. Consequently, I’m not so sure what readers can really learn from this study.

Hide

ED: Publish subject to technical corrections (01 Apr 2015) by James Annan

Short summary

Irrigation agriculture is required to sustain yields that allow feeding the world population. A robust assessment of irrigation requirement (IRR) relies on a sound quantification of evapotranspiration (ET). We prepared a multi-model ensemble considering several ET methods and investigate uncertainties in simulating IRR. More generally, we provide an example of the value of investigating the uncertainty in models that may be used to inform policy-making and to elaborate best management practices.