The Role of Hydrological Signatures in Calibration of Conceptual Hydrological Model
Round 1
Reviewer 1 Report
This paper deals with the role of hydrological signatures in calibration of conceptual hydrological model. The title itself seems illogical because numerical models are usually calibrated. If the results of the model are numerical values of hydrological parameters, then it is a numerical model and such a model is calibrated by different methods so that the obtained values best correspond to the measured values.
Thus, as far as the Bilan model itself is concerned, only its concept is shown in Figure 2 throughout the manuscript and is very sparsely described throughout the chapter on that model.
Rows 84-91: Model parameters should be shown and described in the table.
Chapters on the model calibration methods used are also very scarce. The only metrics that explain the goodness of fitting are two parameters, RMSE and KGE.
Lines 107-109: Chapter 2 states analyzed period from 1981 to 2010 which is about 30 years used in climatological analyzes. However, the period for calibration and evaluation is 1981 to 2016. Why not using the first period 1981-2010? Are those two time periods (1981-1998 and 2001-2016) analyzed for homogeneity in any way? There is also a gap of two years between these two periods. How is this explained?
Line 196: Table 4 is mentioned in text prior mentioning tables 2 and 3. Rearrange this according to the rules of indexing.
Figures 3, 4 and 5 are not completely clear because it is not clear what the values on the ordinate axis mean.
In general, the whole manuscript lacks what the title of the manuscript itself is, and that is the calibration process. The model itself (conceptual or numerical) is not adequately described or demonstrated in practice, and the calibration methods themselves are dryly described in several lines of text.
Therefore, the results shown in the tables are unclear and therefore uninteresting.
I am very sorry, but I cannot give this manuscript any more support, especially in terms of original scientific work. It is recommended that the authors focus more on the calibration methods and the theory behind them and the specific model on which these methods were used. Also that the results of the model and its calibration be more related to some basin characteristics that may affect the quality of model and calibration results. Thus the manuscript is quite diluted with a lot of methods none of which is adequately presented.
Author Response
Summary
We sincerely thank the Editor and the anonymous Reviewers for providing very insightful comments. The article has been corrected based on revisions. Mainly we have made these steps:
- Was added 3 calibrations settings (expanded from 49 to 52) based on discharges quantiles - similar to flow duration curves (FDC) :
- m-daily water: 30, 60, 90, 120, 150, 180, 210,240, 270, 300, 330, 355 and 364 days (time of exceeding of discharges in days in each year),
- m-daily water: 180,
- m-daily water: 300, 330, 355, 364.
- New model run (15 iterations) on these calibration settings (52), without expert calibration.
- New evaluation (mainly based on results GOF) and new selection of the best calibrations settings.
- Was added new scheme of the Bilan model and table which describes parameters, fluxes and water storages of the Bilan model.
- Comparison of FDC’s between manual, automatic, selected and non-selected calibrations settings.
We extended our literature review with a number of relevant references:
References
Westerberg, I. K.; Guerrero, J. L.; Younger, P. M.; Beven, K. J.; Seibert, J.; Halldin, S.; Freer, J. E.; Xu, C. Y. Calibration of hydrological models using flow-duration curves. Hydrology and Earth System Sciences 2011, 2205–2227.
Pfannerstill, M.; Guse, B.; Fohrer, N. Smart low flow signature metrics for an improved overall performance evaluation of hydrological models. Journal of Hydrology 2014, 447–458.
Pokhrel, P.; Yilmaz K.K.; Gupta H.S. Multiple-criteria calibration of a distributed watershed model using spatial regularization and response signatures. Journal of Hydrology 2012, 49–60.
Hrachowitz, M.; O. Fovet, O.; Ruiz; L.; Euser T.; Gharari S.; R. Nijzink R.; J. Freer J.; H. H. G. Savenije, H. H. G.; Gascuel‐Odoux, C. Process consistency in models: The importance of system signatures, expert knowledge, and process complexity. Water resources research 2014, 7445–7469.
Vizina, A.; Hanel M.; Novický, O.; Treml P. Experience from Simulation of Climate Impacts on Water Regime in Monthly and Daily Time Step. Prague, TG Masaryk Water Research Institute 2010.
Machlica, A; Horvát, O.; Horáček, S.; Oosterwijk, J.; Van Loon, A. F.; Fendeková, M.; Van Lanen, H. A. J. Influence of model structure on base flow estimation using Bilan, frier and HBV-light models, Hydrology and Hydromechanics 2012, 242–251.
Tallaksen, L.M.; Van Lanen, H.A.J. Hydrological drought: processes and estimation methods for streamflow and groundwater, Elsevier, 2004.
Zambrano-Bigiarini, Mauricio. HydroGOF: Goodness-of-fit functions for comparison of simulated and observed hydrological time series. R package version 0.4-0, 2020.
Seibert, J. Estimation of parameter uncertainty in the HBV model: Paper presented at the Nordic Hydrological Conference (Akureyri, Iceland-August 1996). Hydrology Research, 1997, 247-262.
Bárdossy, A. Calibration of hydrological model parameters for ungauged catchments. 2007.
Other comments are addressed below.
Reviewer I
This paper deals with the role of hydrological signatures in calibration of conceptual hydrological model. The title itself seems illogical because numerical models are usually calibrated. If the results of the model are numerical values of hydrological parameters, then it is a numerical model and such a model is calibrated by different methods so that the obtained values best correspond to the measured values.
Thus, as far as the Bilan model itself is concerned, only its concept is shown in Figure 2 throughout the manuscript and is very sparsely described throughout the chapter on that model.
Authors' response: The figure 2 was replaced by a more transparent scheme.
Rows 84-91: Model parameters should be shown and described in the table.
Authors' response: The rows 84-91 are introduced in the manuscript as Table 1.
Chapters on the model calibration methods used are also very scarce. The only metrics that explain the goodness of fitting are two parameters, RMSE and KGE.
Authors' response: Is added flow duration curve for better evaluated calibration strategies.
Lines 107-109: Chapter 2 states analyzed period from 1981 to 2010 which is about 30 years used in climatological analyzes. However, the period for calibration and evaluation is 1981 to 2016. Why not using the first period 1981-2010? Are those two time periods (1981-1998 and 2001-2016) analyzed for homogeneity in any way? There is also a gap of two years between these two periods. How is this explained?
Authors' response: The gap of two years between periods was a mistake, actually the period is corrected.
Line 196: Table 4 is mentioned in text prior mentioning tables 2 and 3. Rearrange this according to the rules of indexing.
Authors' response: Is corrected, actually.
Figures 3, 4 and 5 are not completely clear because it is not clear what the values on the ordinate axis mean.
Authors' response: In Figure 3, 4, 5 are described 3 vertical lines, which described: mean selected signatures, mean non-selected signatures, mean signatures which contain time series. The vertical lines represent the difference between selected signatures, non-selected signatures, and time series. The selected signatures were evaluated as appropriate for goodness-of-fit (GOF), parameters of the hydrological model (BP), and high and low flow indices (RS).
Reviewer 2 Report
The research described in the paper is of high interest to potential readers, given the issues of the determination of certain probability discharges have become extremely important nowadays as a consequence of changing time- and spatial distributions of precipitation.
I find the paper interesting and soundly formulated, however I suggest a thorough checking of the text for eventual mistakes in typing and grammar.
Author Response
Summary
We sincerely thank the Editor and the anonymous Reviewers for providing very insightful comments. The article has been corrected based on revisions. Mainly we have made these steps:
- Was added 3 calibrations settings (expanded from 49 to 52) based on discharges quantiles - similar to flow duration curves (FDC) :
- m-daily water: 30, 60, 90, 120, 150, 180, 210,240, 270, 300, 330, 355 and 364 days (time of exceeding of discharges in days in each year),
- m-daily water: 180,
- m-daily water: 300, 330, 355, 364.
- New model run (15 iterations) on these calibration settings (52), without expert calibration.
- New evaluation (mainly based on results GOF) and new selection of the best calibrations settings.
- Was added new scheme of the Bilan model and table which describes parameters, fluxes and water storages of the Bilan model.
- Comparison of FDC’s between manual, automatic, selected and non-selected calibrations settings.
We extended our literature review with a number of relevant references:
References
Westerberg, I. K.; Guerrero, J. L.; Younger, P. M.; Beven, K. J.; Seibert, J.; Halldin, S.; Freer, J. E.; Xu, C. Y. Calibration of hydrological models using flow-duration curves. Hydrology and Earth System Sciences 2011, 2205–2227.
Pfannerstill, M.; Guse, B.; Fohrer, N. Smart low flow signature metrics for an improved overall performance evaluation of hydrological models. Journal of Hydrology 2014, 447–458.
Pokhrel, P.; Yilmaz K.K.; Gupta H.S. Multiple-criteria calibration of a distributed watershed model using spatial regularization and response signatures. Journal of Hydrology 2012, 49–60.
Hrachowitz, M.; O. Fovet, O.; Ruiz; L.; Euser T.; Gharari S.; R. Nijzink R.; J. Freer J.; H. H. G. Savenije, H. H. G.; Gascuel‐Odoux, C. Process consistency in models: The importance of system signatures, expert knowledge, and process complexity. Water resources research 2014, 7445–7469.
Vizina, A.; Hanel M.; Novický, O.; Treml P. Experience from Simulation of Climate Impacts on Water Regime in Monthly and Daily Time Step. Prague, TG Masaryk Water Research Institute 2010.
Machlica, A; Horvát, O.; Horáček, S.; Oosterwijk, J.; Van Loon, A. F.; Fendeková, M.; Van Lanen, H. A. J. Influence of model structure on base flow estimation using Bilan, frier and HBV-light models, Hydrology and Hydromechanics 2012, 242–251.
Tallaksen, L.M.; Van Lanen, H.A.J. Hydrological drought: processes and estimation methods for streamflow and groundwater, Elsevier, 2004.
Zambrano-Bigiarini, Mauricio. HydroGOF: Goodness-of-fit functions for comparison of simulated and observed hydrological time series. R package version 0.4-0, 2020.
Seibert, J. Estimation of parameter uncertainty in the HBV model: Paper presented at the Nordic Hydrological Conference (Akureyri, Iceland-August 1996). Hydrology Research, 1997, 247-262.
Bárdossy, A. Calibration of hydrological model parameters for ungauged catchments. 2007.
Other comments are addressed below.
Reviewer II
The research described in the paper is of high interest to potential readers, given the issues of the determination of certain probability discharges have become extremely important nowadays as a consequence of changing time- and spatial distributions of precipitation.
I find the paper interesting and soundly formulated, however I suggest a thorough checking of the text for eventual mistakes in typing and grammar.
Authors' response: Thank you for your revision, English grammar is fixed.
Reviewer 3 Report
The manuscript „The role of hydrological signatures in calibration of conceptual hydrological model“ by Melisova et al. is about a calibration study for several czech catchments using the Bilan model and different combinations of performance measures and hydrological signatures.
In general, hydrological consistent models are a prerequisite to ensure reliable model results in hydrology. However, the tendency of calibrating a hydrological model with several performance criteria is not new and was described in several other studies before. Consequently, to justify publication of this study, it is necessary that the authors present in detail their new findings and why their study is something new to the hydrological community.
In this regard, I see the need to rework the applied calibration scheme and of course the way comparing the model performance. For example, there are several examples for studies that already discussed the value of calibrating hydrolgical models with a combination of time-series criteria and hydrological signatures. The authors need to explicitely discuss and consider the flow duration curve as an accepted opportunity for model evaluation. Furthermore, it is unclear why the authors do model peformance evaluation leaving out other criteria that are relevant to ensure proper simulation of the hydrologlical processes. Since signatures and components of the hydrological cycle were issued in this manuscript, it would beneficial to evaluate the deviations between measured/estimated hydrololgical components (flow duration curve, ETP, groundwater flow, soil water etc.) for the different calibration strategies. In this regard, a reason for slightly lower model performance for the expert calibration might be a more realistic representation of the hydrological processes, which are not directly considered within the automatic calibration. As stated by the authors, it is reasonable that calibration with time-dependend criteria leads to good performance evaluated with time-dependend perfomance measures. However, it has to be discussed that other important points such as soil water, groundwater, interflow and direct runoff are not directly considered and are therefore likely to be calibrated into the wrong direction. This tends to the issue of „right model results for the wrong reason“.
I strongly recommend to consider the following studies as a starting point for a literature review to finally set up a more modern and established calibration scheme:
https://doi.org/10.5194/hess-15-2205-2011
https://doi.org/10.1002/2014WR016520
https://doi.org/10.1016/j.jhydrol.2013.12.044
https://doi.org/10.1016/j.jhydrol.2008.12.004
https://doi.org/10.1111/1752-1688.12524
https://doi.org/10.1002/2014WR015484
https://doi.org/10.1002/hyp.11300
In my opinion, the authors should apply a model calibration that integrates time-series, several aspects of the flow duration curve and single aspects of the hydrological cycle. Furthermore, evaluation of the different calibration strategies should integrate these three groups of criteria to really see the differences of the hydrological behaviour of the model when it is calibrated with different approaches.
Author Response
Summary
We sincerely thank the Editor and the anonymous Reviewers for providing very insightful comments. The article has been corrected based on revisions. Mainly we have made these steps:
- Was added 3 calibrations settings (expanded from 49 to 52) based on discharges quantiles - similar to flow duration curves (FDC) :
- m-daily water: 30, 60, 90, 120, 150, 180, 210,240, 270, 300, 330, 355 and 364 days (time of exceeding of discharges in days in each year),
- m-daily water: 180,
- m-daily water: 300, 330, 355, 364.
- New model run (15 iterations) on these calibration settings (52), without expert calibration.
- New evaluation (mainly based on results GOF) and new selection of the best calibrations settings.
- Was added new scheme of the Bilan model and table which describes parameters, fluxes and water storages of the Bilan model.
- Comparison of FDC’s between manual, automatic, selected and non-selected calibrations settings.
We extended our literature review with a number of relevant references:
References
Westerberg, I. K.; Guerrero, J. L.; Younger, P. M.; Beven, K. J.; Seibert, J.; Halldin, S.; Freer, J. E.; Xu, C. Y. Calibration of hydrological models using flow-duration curves. Hydrology and Earth System Sciences 2011, 2205–2227.
Pfannerstill, M.; Guse, B.; Fohrer, N. Smart low flow signature metrics for an improved overall performance evaluation of hydrological models. Journal of Hydrology 2014, 447–458.
Pokhrel, P.; Yilmaz K.K.; Gupta H.S. Multiple-criteria calibration of a distributed watershed model using spatial regularization and response signatures. Journal of Hydrology 2012, 49–60.
Hrachowitz, M.; O. Fovet, O.; Ruiz; L.; Euser T.; Gharari S.; R. Nijzink R.; J. Freer J.; H. H. G. Savenije, H. H. G.; Gascuel‐Odoux, C. Process consistency in models: The importance of system signatures, expert knowledge, and process complexity. Water resources research 2014, 7445–7469.
Vizina, A.; Hanel M.; Novický, O.; Treml P. Experience from Simulation of Climate Impacts on Water Regime in Monthly and Daily Time Step. Prague, TG Masaryk Water Research Institute 2010.
Machlica, A; Horvát, O.; Horáček, S.; Oosterwijk, J.; Van Loon, A. F.; Fendeková, M.; Van Lanen, H. A. J. Influence of model structure on base flow estimation using Bilan, frier and HBV-light models, Hydrology and Hydromechanics 2012, 242–251.
Tallaksen, L.M.; Van Lanen, H.A.J. Hydrological drought: processes and estimation methods for streamflow and groundwater, Elsevier, 2004.
Zambrano-Bigiarini, Mauricio. HydroGOF: Goodness-of-fit functions for comparison of simulated and observed hydrological time series. R package version 0.4-0, 2020.
Seibert, J. Estimation of parameter uncertainty in the HBV model: Paper presented at the Nordic Hydrological Conference (Akureyri, Iceland-August 1996). Hydrology Research, 1997, 247-262.
Bárdossy, A. Calibration of hydrological model parameters for ungauged catchments. 2007.
Other comments are addressed below.
Reviewer III
The manuscript „The role of hydrological signatures in calibration of conceptual hydrological model“ by Melisova et al. is about a calibration study for several czech catchments using the Bilan model and different combinations of performance measures and hydrological signatures.
In general, hydrological consistent models are a prerequisite to ensure reliable model results in hydrology. However, the tendency of calibrating a hydrological model with several performance criteria is not new and was described in several other studies before. Consequently, to justify publication of this study, it is necessary that the authors present in detail their new findings and why their study is something new to the hydrological community.
In this regard, I see the need to rework the applied calibration scheme and of course the way comparing the model performance. For example, there are several examples for studies that already discussed the value of calibrating hydrolgical models with a combination of time-series criteria and hydrological signatures. The authors need to explicitely discuss and consider the flow duration curve as an accepted opportunity for model evaluation. Furthermore, it is unclear why the authors do model peformance evaluation leaving out other criteria that are relevant to ensure proper simulation of the hydrologlical processes. Since signatures and components of the hydrological cycle were issued in this manuscript, it would beneficial to evaluate the deviations between measured/estimated hydrololgical components (flow duration curve, ETP, groundwater flow, soil water etc.) for the different calibration strategies. In this regard, a reason for slightly lower model performance for the expert calibration might be a more realistic representation of the hydrological processes, which are not directly considered within the automatic calibration. As stated by the authors, it is reasonable that calibration with time-dependend criteria leads to good performance evaluated with time-dependend perfomance measures. However, it has to be discussed that other important points such as soil water, groundwater, interflow and direct runoff are not directly considered and are therefore likely to be calibrated into the wrong direction. This tends to the issue of „right model results for the wrong reason“.
I strongly recommend to consider the following studies as a starting point for a literature review to finally set up a more modern and established calibration scheme:
https://doi.org/10.5194/hess-15-2205-2011
https://doi.org/10.1002/2014WR016520
https://doi.org/10.1016/j.jhydrol.2013.12.044
https://doi.org/10.1016/j.jhydrol.2008.12.004
https://doi.org/10.1111/1752-1688.12524
https://doi.org/10.1002/2014WR015484
https://doi.org/10.1002/hyp.11300
In my opinion, the authors should apply a model calibration that integrates time-series, several aspects of the flow duration curve and single aspects of the hydrological cycle. Furthermore, evaluation of the different calibration strategies should integrate these three groups of criteria to really see the differences of the hydrological behaviour of the model when it is calibrated with different approaches.
Authors' response: In the article was added all recommended articles to the Introduction. The flow duration curve was added to the Result and discussion. The FDC significantly improved understanding of calibration strategies and their evaluation. Near information about the change is described in the summary.
Round 2
Reviewer 1 Report
The manuscript in this form has been greatly altered and improved over the first version. In the manuscript, it is only necessary to technically improve the graphic representations and correct any typographical and grammatical errors.
Given the new state of the manuscript, I suggest acceptance after minor revision.
Author Response
Dear reviewer,
thank you for your suggestions. In the manuscript are improved and correct any typographical and grammatical errors.
Thank you
Kind regards,
Eva Melišová
Author Response File: 
Author Response.pdf
Reviewer 3 Report
The current manuscript "The role of hydrological signatures in calibration of
conceptual hydrological model" by Melisova et al. is a revision of a previous manuscript version. According to the response letters of the authors and marked changes within the manuscript, it seems that the authors carefully considered the remarks of the reviewers.
With regard to the introduction, I have some minor recommendations:
Line 26: please check grammar and wording
Line 39: additional references should be integrated to fully mention the scientific efforts of other researchers:
https://doi.org/10.1111/1752-1688.12524
https://doi.org/10.5194/hess-22-4425-2018
https://doi.org/10.1016/j.jhydrol.2018.02.055
Line 40: The segmentation of the FDC to successfully calibrate hydrological models was intensively studied in https://doi.org/10.1016/j.jhydrol.2013.12.044 This study is especially useful since the authors mentioned the low flow segment of 95% (firstly applied in this mentioned study) at the discussion and results chapter.
As a final comment, I would like to encourage the authors to consider another crucial point for the conclusions: obviously, automatic calibration leeds to perfect fits between oberserved and simulated discharge. However, with respect to hydrologial realism of the model, it might be more desirable to select a calibrated model with less good performance for simulated discharge at the expense of a model with more reasonable water balance or soil water simulation. Perhaps, the authors should think about a more clear statement about the selecting the option of less model performance in discharge simulation but increased hydrological realism.
Author Response
Dear Reviewer,
thank you for your suggestions.
# line 26: is corrected
# line 39: is corrected
# line 40: is corrected
# the final comments are improved
Thank you
Kind regards,
Eva Melišová
Author Response File: Author Response.pdf
