Producing hydrologic scenarios from raw climate model outputs using an asynchronous modelling framework

Ricard, Simon; Lucas-Picher, Philippe; Anctil, François

doi:https://doi.org/10.5194/hess-2021-451

Preprints

https://doi.org/10.5194/hess-2021-451

Preprints

15 Sep 2021

| 15 Sep 2021

Status: this preprint was under review for the journal HESS but the revision was not accepted.

Producing hydrologic scenarios from raw climate model outputs using an asynchronous modelling framework

Simon Ricard, Philippe Lucas-Picher, and François Anctil

Abstract. Statistical post-processing of climate model outputs is a common hydroclimatic modelling practice aiming to produce climate scenarios that better fit in-situ observations and to produce reliable stream flows forcing calibrated hydrologic models. Such practice is however criticized for disrupting the physical consistency between simulated climate variables and affecting the trends in climate change signals imbedded within raw climate simulations. It also requires abundant good-quality meteorological observations, which are not available for many regions in the world. A simplified hydroclimatic modelling workflow is proposed to quantify the impact of climate change on water discharge without resorting to meteorological observations, nor for statistical post-processing of climate model outputs, nor for calibrating hydrologic models. By combining asynchronous hydroclimatic modelling, an alternative framework designed to construct hydrologic scenarios without resorting to meteorological observations, and quantile perturbation applied to streamflow observations, the proposed workflow produces sound and plausible hydrologic scenarios considering: (1) they preserve trends and physical consistency between simulated climate variables, (2) are implemented from a modelling cascades despite observation scarcity, and (3) support the participation of end-users in producing and interpreting climate change impacts on water resources. The proposed modelling workflow is implemented over four subcatchments of the Chaudière River, Canada, using 9 North American CORDEX simulations and a pool of lumped conceptual hydrologic models. Forced with raw climate model outputs, hydrologic models are calibrated over the reference period according to a calibration metric designed to function with temporally uncorrelated observed and simulated streamflow values. Perturbation factors are defined by relating each simulated streamflow quantiles over both reference and future periods. Hydrologic scenarios are finally produced by applying perturbation factors to available streamflow observations.

Received: 30 Aug 2021 – Discussion started: 15 Sep 2021

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Simon Ricard, Philippe Lucas-Picher, and François Anctil

Status: closed

RC1:
'Comment on hess-2021-451', Anonymous Referee #1, 15 Oct 2021

I am pleased to provide a review of the manuscript “Producing hydrologic scenarios from raw climate model outputs using an asynchronous modelling framework” by Simon Ricard, Philippe Lucas-Picher, and François Anctil. This study attempts to develop a framework to construct hydrologic scenarios without resorting to meteorological observations. The framework has been applied over four subcatchments of the Chaudière River, Canada, using North American CORDEX simulations and a pool of lumped conceptual hydrologic models. The subject is appropriate to be published in the Hydrology and Earth System Sciences. Also, I personally agree with the alternative concept to offset a conventional limitation in the climate change assessment modelling cascade. Although the topic of this study is very interesting, the paper significantly needs to be improved.

To sum up, I recommend a rejection for the current format but encourage resubmitting their draft with a substantial revision. I hope that my comments will be helpful for the authors.

1.       Draft structure: section 2.1 “The Chaudière River catchment” is not appropriate for the section of Methods. It should be separated. Also, section 2.3 “Climate model outputs” is not appropriate for the method section. This draft does not have the results section, which would be proper for section 3. The contents of the draft have to be reorganized.

2.       No method description: The novelty of this draft is to suggest a workflow, enabling the production of streamflow projections without post-processing climate model outputs and without using meteorological observations. However, the draft does not explain their methods in detail. Although the authors do not need to present all the methods that they have used in this draft, they should provide sufficient information related to their novelty. The workflow section (i.e., section 2.2), which, I believe, is the most important section in this draft, mainly declare previous studies to explain their methods. In addition, they have written mainly less than 10 lines in the said section. Their way of describing their methods is not desirable. To be specific, the authors should describe the details of the asynchronous calibration loop and the quantile perturbation in their contents.

3.       L44: insert one more “ intermodel similarities (Ahn et al. 2019)”

“Incorporating climate model similarities and hydrologic error models to quantify climate change impacts on future riverine flood risk”

4.       L118: In Figure 1, it is hard to recognize the number for each site. Increase its size.

5.       L123: Did you use all recorded years for streamflow? if not, please be specific.

6.       Table1: "Hydrometric station" This is not well defined. Is this station ID?

7.       Table1: Area unit: superscript

8.       L202: the same evapotranspiration formulation: what method did you use?

9.       Figure7: What is "P" here?

10.   L304: exclude “never”

11.   L347: asses a typo

Overall, the draft is relatively short. The authors should provide sufficient information for the method and results.

Citation: https://doi.org/10.5194/hess-2021-451-RC1
- AC1:
  'Reply on RC1', Simon Ricard, 24 Oct 2021
  We are pleased to read that reviewer 1 shares our views on the proposed framework. We also acknowledge the atypical structure of the manuscript that may not support the reader while searching for specific information. We are fully open to rewrite the document according to a more conventional structure as advocated.
  More precisely, in the next version of the manuscript, Section 2 will present the domain and the data. Then, Section 3 will gather all information describing the 3 most relevant methodological elements: the modeling workflow, the hydrologic modelling setup, and the construction of hydrologic scenarios using quantile perturbation. An emphasis will put in further describing the modeling workflow, the calibration loop, and quantile perturbation. Finally, Section 4 will be dedicated to presenting the results.
  Regarding specific comments given by reviewer 1, we will:
  Add the proposed reference at L44;
  
  Increase font size in Figure 1;
  
  Explain explicitly the period for which data were used at L123;
  
  Clarify the ‘’ID station’’ notification;
  
  Correct the area unit superscript in Table 1;
  
  Explicitly describe the evapotranspiration formulation in L202;
  
  Explicitly describe P as the cumulative probability in Figure 7.
  
  Citation: https://doi.org/10.5194/hess-2021-451-AC1
RC2:
'Comment on hess-2021-451', Anonymous Referee #2, 15 Oct 2021

The manuscript “Producing hydrological scenarios from raw climate model outputs using an asynchronous modelling framework” demonstrates a new method to simplify the conventional climate-hydrological modelling chain for climate impact assessment. This method basically only needs to calibrate the hydrological models using the raw regional climate model (RCM) outputs in the historical period against the observed empirical cumulative distribution and then run the calibrated hydrological model using the raw RCM outputs in the future periods. Finally, the future hydrological scenarios can be generated based on the observed discharge and the perturbation factors derived from the simulated hydrological response in the historical and future periods.

This method sounds promising, but I am disappointed that the authors did not present a comprehensive validation, which is especially important to introduce a new method. There is only one figure (Fig. 6) to visualize the validation results, but the shift of the spring peaks cannot persuade me to trust the method. Based on the results, I am thinking whether this method can be applied for all RCM outputs or only some RCM outputs that contain insignificant biases. Like other hydrological model calibration studies, the authors should plot the empirical cumulative distribution against the observed one and provide statistical criteria results to evaluate the two different distributions in both calibration and validation periods. In addition, the hydrological scenarios generated by this method should be validated. Therefore, the authors should generate a hydrological “scenario” for the historical period and validate it against observations in terms of various aspects, e.g. extremes, water balance and seasonal dynamics. The above mentioned validation procedure is just to prove that the new method generates reasonable scenarios. It will add more values if the authors can compare the “scenarios” for the historical period generated by both the conventional and new methods with observations.

The structure of this manuscript is also problematic. The title of section 2 is methods, but this section includes study area, data, methods and results. I am afraid that the authors need to re-write the manuscript based on the two major problems, but I am looking forward to seeing comprehensive validation results for the proposed method in future.

Citation: https://doi.org/10.5194/hess-2021-451-RC2
- AC2:
  'Reply on RC2', Simon Ricard, 24 Oct 2021
  
  The validation of the framework proposed in the submitted manuscript will be improved, detailed, and supported by additional results. Reviewer 2 provides sound solutions to enforce the validation framework. A formal comparison of empirical cumulative distributions and additional statistical criteria issued from historical scenarios against observations will be added and discussed as proposed.
  We would argue however that a strict comparison between the reference scenarios and the observations could be hazardous in assessing the confidence attributed to the method. Indeed, this method excludes both the drawbacks of the statistical post processing on the consistency of the simulated climate variables and the scope and finality of the study. To give an example, the trust attributed to a scenario will be different depending if a study focuses on spring floods or water stress. These elements will be fully discussed in the revised manuscript.
  The comment given by reviewer 2 on the structure of the manuscript finally confirms that we must revise the document according to a more conventional form, with sections specifically dedicated to the methods and results (see also the reply to RC1).
  
  Citation: https://doi.org/10.5194/hess-2021-451-AC2
  - RC3: 'Reply on AC2', Anonymous Referee #2, 05 Nov 2021
    
    I am looking forward to seeing the new validation results as the authors replied. Just one clarification about the validation. The validation I mentioned is to run the model with the observed meteorological data and compare the simulated and observed discharges. As I understood, the validation the authors mentioned is to run the model with the reference scenario data. It would be good to see both validation results. For the latter validation, I fully agree that only the seasonal hydrograph or extreme statistics can be compared. Based on the current results, it seems that the model cannot reproduce the seasonal hydrograph well if the climate model outputs have large biases. Hence, I agree that the authors focus on the extreme statistics. Then I would appreciate that the authors can discuss the restrictions or the weakness of this method. For example, this method may be more appropirate for floods or for the climate model outputs that have small biases.
    
    Citation: https://doi.org/10.5194/hess-2021-451-RC3
    
    AC3: 'Reply on RC3', Simon Ricard, 14 Nov 2021
    
    Thank you very much for the clarification. We intend to address both validation schemes focusing on seasonal extreme values: 1) a split sample test running hydrologic models with reference scenario data and 2) comparing simulated and observed discharges, running the model with the observed meteorological data. The validation results will be discussed in relation to the applicability and restrictions of the proposed framework.
    
    Citation: https://doi.org/10.5194/hess-2021-451-AC3

Status: closed

RC1:
'Comment on hess-2021-451', Anonymous Referee #1, 15 Oct 2021

I am pleased to provide a review of the manuscript “Producing hydrologic scenarios from raw climate model outputs using an asynchronous modelling framework” by Simon Ricard, Philippe Lucas-Picher, and François Anctil. This study attempts to develop a framework to construct hydrologic scenarios without resorting to meteorological observations. The framework has been applied over four subcatchments of the Chaudière River, Canada, using North American CORDEX simulations and a pool of lumped conceptual hydrologic models. The subject is appropriate to be published in the Hydrology and Earth System Sciences. Also, I personally agree with the alternative concept to offset a conventional limitation in the climate change assessment modelling cascade. Although the topic of this study is very interesting, the paper significantly needs to be improved.

To sum up, I recommend a rejection for the current format but encourage resubmitting their draft with a substantial revision. I hope that my comments will be helpful for the authors.

1.       Draft structure: section 2.1 “The Chaudière River catchment” is not appropriate for the section of Methods. It should be separated. Also, section 2.3 “Climate model outputs” is not appropriate for the method section. This draft does not have the results section, which would be proper for section 3. The contents of the draft have to be reorganized.

2.       No method description: The novelty of this draft is to suggest a workflow, enabling the production of streamflow projections without post-processing climate model outputs and without using meteorological observations. However, the draft does not explain their methods in detail. Although the authors do not need to present all the methods that they have used in this draft, they should provide sufficient information related to their novelty. The workflow section (i.e., section 2.2), which, I believe, is the most important section in this draft, mainly declare previous studies to explain their methods. In addition, they have written mainly less than 10 lines in the said section. Their way of describing their methods is not desirable. To be specific, the authors should describe the details of the asynchronous calibration loop and the quantile perturbation in their contents.

3.       L44: insert one more “ intermodel similarities (Ahn et al. 2019)”

“Incorporating climate model similarities and hydrologic error models to quantify climate change impacts on future riverine flood risk”

4.       L118: In Figure 1, it is hard to recognize the number for each site. Increase its size.

5.       L123: Did you use all recorded years for streamflow? if not, please be specific.

6.       Table1: "Hydrometric station" This is not well defined. Is this station ID?

7.       Table1: Area unit: superscript

8.       L202: the same evapotranspiration formulation: what method did you use?

9.       Figure7: What is "P" here?

10.   L304: exclude “never”

11.   L347: asses a typo

Overall, the draft is relatively short. The authors should provide sufficient information for the method and results.

Citation: https://doi.org/10.5194/hess-2021-451-RC1
- AC1:
  'Reply on RC1', Simon Ricard, 24 Oct 2021
  We are pleased to read that reviewer 1 shares our views on the proposed framework. We also acknowledge the atypical structure of the manuscript that may not support the reader while searching for specific information. We are fully open to rewrite the document according to a more conventional structure as advocated.
  More precisely, in the next version of the manuscript, Section 2 will present the domain and the data. Then, Section 3 will gather all information describing the 3 most relevant methodological elements: the modeling workflow, the hydrologic modelling setup, and the construction of hydrologic scenarios using quantile perturbation. An emphasis will put in further describing the modeling workflow, the calibration loop, and quantile perturbation. Finally, Section 4 will be dedicated to presenting the results.
  Regarding specific comments given by reviewer 1, we will:
  Add the proposed reference at L44;
  
  Increase font size in Figure 1;
  
  Explain explicitly the period for which data were used at L123;
  
  Clarify the ‘’ID station’’ notification;
  
  Correct the area unit superscript in Table 1;
  
  Explicitly describe the evapotranspiration formulation in L202;
  
  Explicitly describe P as the cumulative probability in Figure 7.
  
  Citation: https://doi.org/10.5194/hess-2021-451-AC1
RC2:
'Comment on hess-2021-451', Anonymous Referee #2, 15 Oct 2021

The manuscript “Producing hydrological scenarios from raw climate model outputs using an asynchronous modelling framework” demonstrates a new method to simplify the conventional climate-hydrological modelling chain for climate impact assessment. This method basically only needs to calibrate the hydrological models using the raw regional climate model (RCM) outputs in the historical period against the observed empirical cumulative distribution and then run the calibrated hydrological model using the raw RCM outputs in the future periods. Finally, the future hydrological scenarios can be generated based on the observed discharge and the perturbation factors derived from the simulated hydrological response in the historical and future periods.

This method sounds promising, but I am disappointed that the authors did not present a comprehensive validation, which is especially important to introduce a new method. There is only one figure (Fig. 6) to visualize the validation results, but the shift of the spring peaks cannot persuade me to trust the method. Based on the results, I am thinking whether this method can be applied for all RCM outputs or only some RCM outputs that contain insignificant biases. Like other hydrological model calibration studies, the authors should plot the empirical cumulative distribution against the observed one and provide statistical criteria results to evaluate the two different distributions in both calibration and validation periods. In addition, the hydrological scenarios generated by this method should be validated. Therefore, the authors should generate a hydrological “scenario” for the historical period and validate it against observations in terms of various aspects, e.g. extremes, water balance and seasonal dynamics. The above mentioned validation procedure is just to prove that the new method generates reasonable scenarios. It will add more values if the authors can compare the “scenarios” for the historical period generated by both the conventional and new methods with observations.

The structure of this manuscript is also problematic. The title of section 2 is methods, but this section includes study area, data, methods and results. I am afraid that the authors need to re-write the manuscript based on the two major problems, but I am looking forward to seeing comprehensive validation results for the proposed method in future.

Citation: https://doi.org/10.5194/hess-2021-451-RC2
- AC2:
  'Reply on RC2', Simon Ricard, 24 Oct 2021
  
  The validation of the framework proposed in the submitted manuscript will be improved, detailed, and supported by additional results. Reviewer 2 provides sound solutions to enforce the validation framework. A formal comparison of empirical cumulative distributions and additional statistical criteria issued from historical scenarios against observations will be added and discussed as proposed.
  We would argue however that a strict comparison between the reference scenarios and the observations could be hazardous in assessing the confidence attributed to the method. Indeed, this method excludes both the drawbacks of the statistical post processing on the consistency of the simulated climate variables and the scope and finality of the study. To give an example, the trust attributed to a scenario will be different depending if a study focuses on spring floods or water stress. These elements will be fully discussed in the revised manuscript.
  The comment given by reviewer 2 on the structure of the manuscript finally confirms that we must revise the document according to a more conventional form, with sections specifically dedicated to the methods and results (see also the reply to RC1).
  
  Citation: https://doi.org/10.5194/hess-2021-451-AC2
  - RC3: 'Reply on AC2', Anonymous Referee #2, 05 Nov 2021
    
    I am looking forward to seeing the new validation results as the authors replied. Just one clarification about the validation. The validation I mentioned is to run the model with the observed meteorological data and compare the simulated and observed discharges. As I understood, the validation the authors mentioned is to run the model with the reference scenario data. It would be good to see both validation results. For the latter validation, I fully agree that only the seasonal hydrograph or extreme statistics can be compared. Based on the current results, it seems that the model cannot reproduce the seasonal hydrograph well if the climate model outputs have large biases. Hence, I agree that the authors focus on the extreme statistics. Then I would appreciate that the authors can discuss the restrictions or the weakness of this method. For example, this method may be more appropirate for floods or for the climate model outputs that have small biases.
    
    Citation: https://doi.org/10.5194/hess-2021-451-RC3
    
    AC3: 'Reply on RC3', Simon Ricard, 14 Nov 2021
    
    Thank you very much for the clarification. We intend to address both validation schemes focusing on seasonal extreme values: 1) a split sample test running hydrologic models with reference scenario data and 2) comparing simulated and observed discharges, running the model with the observed meteorological data. The validation results will be discussed in relation to the applicability and restrictions of the proposed framework.
    
    Citation: https://doi.org/10.5194/hess-2021-451-AC3

Simon Ricard, Philippe Lucas-Picher, and François Anctil

Viewed

Total article views: 1,331 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
940	338	53	1,331	43	43

HTML: 940
PDF: 338
XML: 53
Total: 1,331
BibTeX: 43
EndNote: 43

Views and downloads (calculated since 15 Sep 2021)

Month	HTML	PDF	XML	Total
Sep 2021	190	48	4	242
Oct 2021	128	31	8	167
Nov 2021	96	32	2	130
Dec 2021	40	11	1	52
Jan 2022	36	6	0	42
Feb 2022	21	9	0	30
Mar 2022	21	7	1	29
Apr 2022	18	9	0	27
May 2022	13	9	1	23
Jun 2022	3	1	1	5
Jul 2022	24	5	0	29
Aug 2022	9	6	0	15
Sep 2022	7	7	0	14
Oct 2022	17	1	4	22
Nov 2022	19	5	0	24
Dec 2022	16	3	0	19
Jan 2023	19	16	0	35
Feb 2023	7	4	0	11
Mar 2023	15	4	2	21
Apr 2023	6	5	0	11
May 2023	3	2	0	5
Jun 2023	10	5	2	17
Jul 2023	28	9	3	40
Aug 2023	14	7	1	22
Sep 2023	11	10	2	23
Oct 2023	13	5	4	22
Nov 2023	9	3	0	12
Dec 2023	8	1	1	10
Jan 2024	5	6	0	11
Feb 2024	8	15	3	26
Mar 2024	12	20	1	33
Apr 2024	13	4	5	22
May 2024	7	3	2	12
Jun 2024	5	6	1	12
Jul 2024	9	2	3	14
Aug 2024	16	3	0	19
Sep 2024	12	2	0	14
Oct 2024	15	9	0	24
Nov 2024	28	7	1	36
Dec 2024	9	0	9

Cumulative views and downloads (calculated since 15 Sep 2021)

Month	HTML	PDF	XML	Total
Sep 2021	190	48	4	242
Oct 2021	128	31	8	167
Nov 2021	96	32	2	130
Dec 2021	40	11	1	52
Jan 2022	36	6	0	42
Feb 2022	21	9	0	30
Mar 2022	21	7	1	29
Apr 2022	18	9	0	27
May 2022	13	9	1	23
Jun 2022	3	1	1	5
Jul 2022	24	5	0	29
Aug 2022	9	6	0	15
Sep 2022	7	7	0	14
Oct 2022	17	1	4	22
Nov 2022	19	5	0	24
Dec 2022	16	3	0	19
Jan 2023	19	16	0	35
Feb 2023	7	4	0	11
Mar 2023	15	4	2	21
Apr 2023	6	5	0	11
May 2023	3	2	0	5
Jun 2023	10	5	2	17
Jul 2023	28	9	3	40
Aug 2023	14	7	1	22
Sep 2023	11	10	2	23
Oct 2023	13	5	4	22
Nov 2023	9	3	0	12
Dec 2023	8	1	1	10
Jan 2024	5	6	0	11
Feb 2024	8	15	3	26
Mar 2024	12	20	1	33
Apr 2024	13	4	5	22
May 2024	7	3	2	12
Jun 2024	5	6	1	12
Jul 2024	9	2	3	14
Aug 2024	16	3	0	19
Sep 2024	12	2	0	14
Oct 2024	15	9	0	24
Nov 2024	28	7	1	36
Dec 2024	9	0	9

Viewed (geographical distribution)

Total article views: 1,210 (including HTML, PDF, and XML) Thereof 1,210 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 13 Dec 2024

Short summary

We propose a simplified hydroclimatic modelling workflow for producing hydrologic scenarios without resorting to meteorological observations. This innovative approach preserves trends and physical consistency between simulated climate variables, allows the implementation of modelling cascades despite observation scarcity, and supports the participation of end-users in producing and interpreting climate change impacts on water resources.


Total:	0
HTML:	0
PDF:	0
XML:	0