Improving Runoff Simulation in the Western United States with Noah-MP and VIC

Su, Lu; Lettenmaier, Dennis P.; Pan, Ming; Bass, Benjamin

doi:https://doi.org/10.5194/egusphere-2023-2164

Preprints

https://doi.org/10.5194/egusphere-2023-2164

Preprints

28 Sep 2023

| 28 Sep 2023

Improving Runoff Simulation in the Western United States with Noah-MP and VIC

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Abstract. Streamflow forecasts are critical for water and environmental management, especially in the water-short Western U.S.. Land Surface Models (LSMs), such as the Variable Infiltration Capacity (VIC) model and the Noah-Multiparameterization (Noah-MP) play an essential role in providing comprehensive runoff forecasts across the region. Virtually all LSMs require parameter estimation to optimize their predictive capabilities. We describe a systematic calibration of parameters for VIC and Noah-MP over 263 river basins in the Western U.S., and distribution of the calibrated parameters over the entire region. Post-calibration results showed a notable improvement in model accuracy in the calibration basins: the median daily streamflow Kling-Gupta Efficiency (KGE) for VIC rose from 0.37 to 0.70, and for Noah-MP, from 0.22 to 0.54. Employing the donor-basin regionalization method, we developed transfer relationships to hydrologically similar basins and extended the calibrated parameters to ungauged basins and the entire region. We assessed factors that influence calibration efficiency and model performance using regional parameter estimates. We evaluated high and low flow simulation capabilities of the two models and observed marked improvements after calibration and regionalization. We also generated gridded parameter sets for both models across all 4816 HUC-10 basins in the Western U.S., a data set that is intended to support regional hydrologic studies and hydrologic climate change assessments.

Received: 21 Sep 2023 – Discussion started: 28 Sep 2023

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

Preprint (PDF, 3033 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (3033 KB)

Supplement (2034 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

16 Jul 2024

Improving runoff simulation in the Western United States with Noah-MP and variable infiltration capacity

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Hydrol. Earth Syst. Sci., 28, 3079–3097, https://doi.org/10.5194/hess-28-3079-2024,https://doi.org/10.5194/hess-28-3079-2024, 2024

Short summary

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Interactive discussion

Status: closed

CC1:
'A student review on egusphere-2023-2164', Adriaan J. (Ryan) Teuling, 05 Nov 2023

This review was prepared by one of the students at Wageningen University as part of graduate program course work, and has been produced under supervision of Ryan Teuling. The review has been posted because of its good quality, and likely usefulness to the authors and editor. Please use to your benefit.

Citation: https://doi.org/10.5194/egusphere-2023-2164-CC1
- AC3: 'Reply on CC1', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  Thank you for your detailed comments and suggestions. We have taken careful consideration of each point you raised and please refer to the attached document for the responses to your comments.
  Warm regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC3
RC1:
'Comment on egusphere-2023-2164', Anonymous Referee #1, 06 Nov 2023

I enjoyed reading this fruitful manuscript. I only have several concerns and comments listed below.

-L186: the readers would except a similar treatment of the models for consistency of the framework. Except for the objective function (KGE), all other experimental details are different. Different search algorithms (SCE vs DDS), different max iterations (may be understandable). The chaotic structure of the paper makes is difficult to enjoy the fruitful outcomes. Currently it reads like a technical report by the corps engineers doing everything for better KGE. However, the research design needs a framework. The experiments should focus on effects of one thing one at a time. Using two different models is important for structural uncertainty; however under the same P, PET, Tavg input and methodology (calibration algorithm, max iterations etc).

-Please a conceptual diagram/framework describing your methodology.

-Metric: Both models are top quality spatially semi or fully distributed models. Using a metric focusing on average flows may not fit best while NSE focuses more on high flows. There are other metrics focusing on patterns of simulated flux maps such as SSIM, FSS, EOF, SPAEF and SPEM. For novelty, it is recommended to elaborate calibration approach for the two models.

-Did authors applied a sensitivity analysis before the calibration to select most important parameters for the streamflow using selected metric, KGE?

-Figure4: For process consistency, different remote sensing products could be used to evaluate model results. Why did you only focus on streamflows, if the models are capable of producing AET, SM (at different soil horizons). MODIS, SMAP, SMOS, ESA, ALEXI all are useful constraints for model calibration making the novelty maximized. Literature review of the paper misses all those calibration papers focusing not only KGE but also RS products.

https://doi.org/10.1002/2017WR021346

https://doi.org/10.5194/hess-22-1299-2018

-Do these models incorporate pedo-transfer functions for parameter regionalization? Some of the distributed models, like mHM, uses multi-parameter regionalization (MPR) technology.
https://doi.org/10.5194/gmd-15-859-2022

-Table 1: “VIC4.1.2”
Why this old version of VIC model is used while current version WRF-Hydro 5.2.0 is preferred.

VIC5 version includes many infrastructure improvements (glaciers etc) as described here:
https://doi.org/10.5194/gmd-11-3481-2018

-Figures 9-10-11 can be given in appendix.

-The paper needs a separate Discussion section and a separate Conclusions (bullets) section. Summary can be appropriate for “engineering corps” reports not for HESS papers.

Citation: https://doi.org/10.5194/egusphere-2023-2164-RC1
- AC1: 'Reply on RC1', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  Thank you for your insightful feedback on our manuscript. For your convenience, we have detailed our responses in the attached document.
  We are grateful for the time and effort you have invested in reviewing our work .
  Best regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC1
RC2:
'Comment on egusphere-2023-2164', Anonymous Referee #2, 06 Nov 2023

General comments

This study presents the calibration and regionalisation of two land-surface hydrologic models in 263 catchments in the Western US. The results indicate that the median Kling-Gupta efficiency obtained in model calibration and regionalisation outperforms earlier/baseline study ().

The study focuses on an interesting topic, but in its current form, it reads more like a technical report than a research paper. The Introduction nicely presents the context (i.e. why it is important to simulate water cycle components in the study region accurately). Still, the synthesis and formulation of the current research gaps need to be significantly improved. There is a large body of literature focusing on regional calibration of hydrologic models, transfer of model parameters at the regional scale, definition of the signatures used for similarity definition etc. It needs to be made clear how this study goes beyond the existing studies. The formulation of the research questions needs to be very precise and linked with presenting the research gaps. In its current form, it needs to be clarified whether the main goal is to propose and evaluate some methodological advance in model calibration and/or regionalisation or to present some new factual information about the study region (a case study analysis).

The selection of the two models needs to be better justified. What are the differences in runoff generation between the models (and how is it linked with the regional variability of runoff generation in the study region)? It needs to be clarified why to use a 3hr simulation time step, when model inputs are daily. It is also not clear why to calibrate only selected soil-related parameters and how the selection is linked with the runoff generation processes and their variability in the study region. For example, are the snow accumulation and melt processes less important? Or are the snow-related model parameters already accurately calibrated? More importantly, the results and the differences between the two models need to be better linked with the main runoff generation processes (and their regional variability).

I missed the discussion of the results, which will link the new findings with previous studies. This can enhance the demonstration of the novel scientific contribution of the study.

Citation: https://doi.org/10.5194/egusphere-2023-2164-RC2
- AC2: 'Reply on RC2', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  We sincerely thank you for your valuable comments on our manuscript. Each of your points has been thoughtfully considered and addressed, as outlined in the attached response.
  Warm regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC2

Interactive discussion

Status: closed

CC1:
'A student review on egusphere-2023-2164', Adriaan J. (Ryan) Teuling, 05 Nov 2023

This review was prepared by one of the students at Wageningen University as part of graduate program course work, and has been produced under supervision of Ryan Teuling. The review has been posted because of its good quality, and likely usefulness to the authors and editor. Please use to your benefit.

Citation: https://doi.org/10.5194/egusphere-2023-2164-CC1
- AC3: 'Reply on CC1', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  Thank you for your detailed comments and suggestions. We have taken careful consideration of each point you raised and please refer to the attached document for the responses to your comments.
  Warm regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC3
RC1:
'Comment on egusphere-2023-2164', Anonymous Referee #1, 06 Nov 2023

I enjoyed reading this fruitful manuscript. I only have several concerns and comments listed below.

-L186: the readers would except a similar treatment of the models for consistency of the framework. Except for the objective function (KGE), all other experimental details are different. Different search algorithms (SCE vs DDS), different max iterations (may be understandable). The chaotic structure of the paper makes is difficult to enjoy the fruitful outcomes. Currently it reads like a technical report by the corps engineers doing everything for better KGE. However, the research design needs a framework. The experiments should focus on effects of one thing one at a time. Using two different models is important for structural uncertainty; however under the same P, PET, Tavg input and methodology (calibration algorithm, max iterations etc).

-Please a conceptual diagram/framework describing your methodology.

-Metric: Both models are top quality spatially semi or fully distributed models. Using a metric focusing on average flows may not fit best while NSE focuses more on high flows. There are other metrics focusing on patterns of simulated flux maps such as SSIM, FSS, EOF, SPAEF and SPEM. For novelty, it is recommended to elaborate calibration approach for the two models.

-Did authors applied a sensitivity analysis before the calibration to select most important parameters for the streamflow using selected metric, KGE?

-Figure4: For process consistency, different remote sensing products could be used to evaluate model results. Why did you only focus on streamflows, if the models are capable of producing AET, SM (at different soil horizons). MODIS, SMAP, SMOS, ESA, ALEXI all are useful constraints for model calibration making the novelty maximized. Literature review of the paper misses all those calibration papers focusing not only KGE but also RS products.

https://doi.org/10.1002/2017WR021346

https://doi.org/10.5194/hess-22-1299-2018

-Do these models incorporate pedo-transfer functions for parameter regionalization? Some of the distributed models, like mHM, uses multi-parameter regionalization (MPR) technology.
https://doi.org/10.5194/gmd-15-859-2022

-Table 1: “VIC4.1.2”
Why this old version of VIC model is used while current version WRF-Hydro 5.2.0 is preferred.

VIC5 version includes many infrastructure improvements (glaciers etc) as described here:
https://doi.org/10.5194/gmd-11-3481-2018

-Figures 9-10-11 can be given in appendix.

-The paper needs a separate Discussion section and a separate Conclusions (bullets) section. Summary can be appropriate for “engineering corps” reports not for HESS papers.

Citation: https://doi.org/10.5194/egusphere-2023-2164-RC1
- AC1: 'Reply on RC1', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  Thank you for your insightful feedback on our manuscript. For your convenience, we have detailed our responses in the attached document.
  We are grateful for the time and effort you have invested in reviewing our work .
  Best regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC1
RC2:
'Comment on egusphere-2023-2164', Anonymous Referee #2, 06 Nov 2023

General comments

This study presents the calibration and regionalisation of two land-surface hydrologic models in 263 catchments in the Western US. The results indicate that the median Kling-Gupta efficiency obtained in model calibration and regionalisation outperforms earlier/baseline study ().

The study focuses on an interesting topic, but in its current form, it reads more like a technical report than a research paper. The Introduction nicely presents the context (i.e. why it is important to simulate water cycle components in the study region accurately). Still, the synthesis and formulation of the current research gaps need to be significantly improved. There is a large body of literature focusing on regional calibration of hydrologic models, transfer of model parameters at the regional scale, definition of the signatures used for similarity definition etc. It needs to be made clear how this study goes beyond the existing studies. The formulation of the research questions needs to be very precise and linked with presenting the research gaps. In its current form, it needs to be clarified whether the main goal is to propose and evaluate some methodological advance in model calibration and/or regionalisation or to present some new factual information about the study region (a case study analysis).

The selection of the two models needs to be better justified. What are the differences in runoff generation between the models (and how is it linked with the regional variability of runoff generation in the study region)? It needs to be clarified why to use a 3hr simulation time step, when model inputs are daily. It is also not clear why to calibrate only selected soil-related parameters and how the selection is linked with the runoff generation processes and their variability in the study region. For example, are the snow accumulation and melt processes less important? Or are the snow-related model parameters already accurately calibrated? More importantly, the results and the differences between the two models need to be better linked with the main runoff generation processes (and their regional variability).

I missed the discussion of the results, which will link the new findings with previous studies. This can enhance the demonstration of the novel scientific contribution of the study.

Citation: https://doi.org/10.5194/egusphere-2023-2164-RC2
- AC2: 'Reply on RC2', Lu Su, 21 Dec 2023
  
  Dear Reviewer,
  We sincerely thank you for your valuable comments on our manuscript. Each of your points has been thoughtfully considered and addressed, as outlined in the attached response.
  Warm regards,
  Lu
  
  Citation: https://doi.org/10.5194/egusphere-2023-2164-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (08 Jan 2024) by Markus Hrachowitz

Dear authors,

as you have seen we have received detailed and very constructive assessments of your manuscript from two reviewers and one additional community comment.
Overall, the objective and intention of your work has been positively received. However, the reviewers also flag a number of points that need to be adequately addressed before the manuscript can be considered for publication. From my perspective, in particular the following points will warrant some more attention:

(1) As pointed out by both reviewers, the manuscript reads more like a technical report than a scientific paper. This aspect has to be substantially developed. In particular, it remains unclear what the novelty of the analysis is. In addition, it will be important to considerably expand on what can be learned from (or not) your experiment and what the implications thereof are. In the end, one of the objectives of a well-developed scientific paper is to pro-actively avoid the reader wanting to ask: "So what?"

(2) The rationale behind the choice of using these two models (and not others) will benefit from being strengthened as well.

(3) I also agree with the reviewers that the choice to limit calibration efforts to time series of stream flow and one single objective function comes a bit surprising in the light of the increasing body of literature over the past one or two decades that illustrates the benefit and even the necessity to develop much broader model calibration/evaluation schemes, either using multiple different objective functions (including spatial pattern) and/or calibration variables (e.g. hydrological signatures or other observable variables, e.g. snow cover).

(4) it remains completely unclear what the benefit of running the model at a sub-daily time-scale is, when the input data are daily. How can this be meaningfully done? What is the benefit? How are modelling artifacts due to averaging effects in the daily data avoided at the sub-daily time scale?

On balance, this manuscript can have some potential. However the above points and the remaining reviewer comments need to be addressed in detail and convincingly, which may require some substantial re-thinking and re-working of the experiment.
However, I believe this could be possible in a round of quite major revisions.

I am looking forward to receiving a revised version of your manuscript.
Best regards,
Markus Hrachowitz

Hide

AR by Lu Su on behalf of the Authors (16 Feb 2024) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (16 Feb 2024) by Markus Hrachowitz

RR by Anonymous Referee #1 (16 Feb 2024)

RR by Anonymous Referee #2 (27 Feb 2024)

Suggestions for revision or reasons for rejection

Dear editor, dear authors,
I’m sorry to say that my concerns and critical comments (from the first review round) have only been partially addressed. My first comment was related to the formulation of the research gaps and the scientific novelty of the research objectives. This part still needs to be substantially improved. The revision includes two new paragraphs indicating that one of the new contributions is the application of two models addressing structural model uncertainty. I wonder to what extent and what precisely is a new contribution here. There are numerous papers examining the uncertainty and its reduction using multi-model ensembles (e.g. Her et al, 2019 among many others), so it is important to review existing studies on this topic along with the formulation of what new research hypothesis is tested in the manuscript. The second new paragraph indicates that the selected models are widely accepted models. I agree. Still there are again numerous comparisons that include these models (e.g. Liu et al, 2020 among many others), so it is important to highlight what are still the open research gaps and what new understanding can be obtained in the current study. From the revisions made, I still do not clearly see what the research gaps are and how the current study adds to a new scientific understanding/knowledge. The calibration of a model is more a technical question than a scientific objective of a research paper.
A large part of the results refers to the regionalisation. The scientific objective of this part (again, research gaps and novelty) are not covered in the Introduction in its current form. There are studies linking the performance of regionalisation methods to the climate or physiographic attributes so it will be interesting to link the current objectives and results with previous studies. Moreover this part seems to be more larger and significant than the calibration part, but many results are just presented in the supplement.
The final remark is related to the interpretation of the correlations between calibrated model efficiency and physiographic attributes. In my opinion most of the correlations are rather weak (even statistically significant) and some more detailed attribution to the processes (and how these are represented in the models) will be an interesting extension of the analysis.

References
Her, Y., Yoo, SH., Cho, J. et al. Uncertainty in hydrological analysis of climate change: multi-parameter vs. multi-GCM ensemble predictions. Sci Rep 9, 4974 (2019). https://doi.org/10.1038/s41598-019-41334-7
Liu et al (2020) https://doi.org/10.1016/j.jhydrol.2019.124534

Hide

ED: Reconsider after major revisions (further review by editor and referees) (27 Feb 2024) by Markus Hrachowitz

AR by Lu Su on behalf of the Authors (03 May 2024) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (08 May 2024) by Markus Hrachowitz

RR by Juraj Parajka (10 May 2024)

ED: Publish as is (21 May 2024) by Markus Hrachowitz

AR by Lu Su on behalf of the Authors (21 May 2024) Manuscript

Journal article(s) based on this preprint

16 Jul 2024

Improving runoff simulation in the Western United States with Noah-MP and variable infiltration capacity

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Hydrol. Earth Syst. Sci., 28, 3079–3097, https://doi.org/10.5194/hess-28-3079-2024,https://doi.org/10.5194/hess-28-3079-2024, 2024

Short summary

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Supplement

https://doi.org/10.5194/egusphere-2023-2164-supplement

Lu Su, Dennis P. Lettenmaier, Ming Pan, and Benjamin Bass

Viewed

Total article views: 780 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
538	202	40	780	55	19	21

HTML: 538
PDF: 202
XML: 40
Total: 780
Supplement: 55
BibTeX: 19
EndNote: 21

Views and downloads (calculated since 28 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	89	20	4	113
Oct 2023	176	66	4	246
Nov 2023	73	26	8	107
Dec 2023	30	13	8	51
Jan 2024	29	7	2	38
Feb 2024	26	14	2	42
Mar 2024	23	16	1	40
Apr 2024	46	10	4	60
May 2024	12	12	2	26
Jun 2024	21	9	4	34
Jul 2024	13	9	1	23

Cumulative views and downloads (calculated since 28 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	89	20	4	113
Oct 2023	176	66	4	246
Nov 2023	73	26	8	107
Dec 2023	30	13	8	51
Jan 2024	29	7	2	38
Feb 2024	26	14	2	42
Mar 2024	23	16	1	40
Apr 2024	46	10	4	60
May 2024	12	12	2	26
Jun 2024	21	9	4	34
Jul 2024	13	9	1	23

Viewed (geographical distribution)

Total article views: 769 (including HTML, PDF, and XML) Thereof 769 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 16 Jul 2024

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (3033 KB)
Metadata XML

Short summary

We used a systematic method to fine-tune two widely used models in 263 river basins in the Western U.S. Then, we developed transfer relationships to similar basins and extended the fine-tuned parameters to ungauged basins . We assessed factors that influence the performance of the parameter estimation and extension. We also checked how well the two models could simulate high and low river flows and saw notable enhancements. This helps studies on regional river flow simulation and forecast.


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