Use of satellite remote sensing to validate reservoir operations in global hydrological models: a case study from the CONUS

Otta, Kedar; Müller Schmied, Hannes; Gosling, Simon N.; Hanasaki, Naota

doi:https://doi.org/10.5194/hess-2023-215

Preprints

https://doi.org/10.5194/hess-2023-215

Preprints

12 Sep 2023

| 12 Sep 2023

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

Use of satellite remote sensing to validate reservoir operations in global hydrological models: a case study from the CONUS

Kedar Otta, Hannes Müller Schmied, Simon N. Gosling, and Naota Hanasaki

Abstract. Although river discharge simulations from global hydrological models have undergone extensive validation, there has been less validation of reservoir operations, primarily because of limited observational data. However, recent advancements in satellite remote sensing technology have facilitated the collection of valuable data regarding water surface area and elevation, thereby providing the ability to validate reservoir storage. In this study, we sought to establish a methodology for validation and intercomparison of reservoir storage within global hydrological model simulations using satellite-derived data. Accordingly, we chose two satellite-derived reservoir operation products, DAHITI and GRSAD, to create monthly time series storage data for seven reservoirs in the contiguous United States (CONUS) , with access to long-term ground truth data (the total catchment area accounts for about 9 % of CONUS). We assessed two global hydrological models that participated in the Inter Sectoral Model Intercomparison Project (ISIMIP) Phase 3 project, H08 and WaterGAP2, with three distinct forcing datasets: GSWP3-W5E5 (GW), CR20v3-W5E5 (CW), and CR20v3-ERA5 (CE). The results indicated that WaterGAP2 generally outperforms H08; the CW forcing dataset demonstrated superior results compared with GW and CE; the DAHITI showed better consistency with ground observations than GRSAD if temporal coverage is sufficient. Overall, our study emphasizes the potential uses of satellite remote sensing data in reservoir operations validation and underscores the importance of normalization and decomposition techniques for improved validation efficacy. The results highlight the relative performances of different hydrological models and forcing datasets, yielding insights concerning future advancements in reservoir simulation and operational studies.

Received: 02 Sep 2023 – Discussion started: 12 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, 2567 KB)

Supplement (2055 KB)

Download & links

Kedar Otta, Hannes Müller Schmied, Simon N. Gosling, and Naota Hanasaki

Status: closed

RC1:
'Comment on hess-2023-215', Anonymous Referee #1, 12 Sep 2023

This is an interesting study. The authors are exploring the use of satellite remote sensing to validate reservoir operations in global hydrological models (GHMs). I think if we take the perspective of the GHMs and their need to be improved as a modeling framework by moving away from assumptions on reservoir operations and parameterizations (which GHMs have a lot of), then this is a valuable study. However, if this study is just exploring if we can use satellite data to track reservoirs, I don’t think this study is innovative or adds much to the body of knowledge. That topic has been addressed and most of the relevant questions on that that the authors pose in this study have been answered much more vigorously and in more detail over the last 10 years. So the first question posed by the authors is really redundant here (we know the answer): Can satellite-based storage estimation data serve as a surrogate for ground truth data?
The authors have used too simplistic methods of using only altimeter from DAHITI (with limited sampling in space and time) and Landsat (optical) based reservoir area dataset from GRSADs. Today when we have a fleet of satellite sensors (optical, SAR, altimeters) that collectively provide more robust, high frequency and more accurate tracking of reservoirs in terms of surface area or elevation to generate storage change. So I think the authors should reorganize the paper and reduce section 3.1 and focus more on 3.2 as the key focus of the paper where there is a lot to learn for the GHM community trying to improve representation of reservoirs on global models (which also fits perfectly into the theme of the special issue of HESS).
Let me share some key references the authors miss in capturing the state of the art of satellite-based reservoir tracking:
Bonnema, M., C.H. David, R.P. d. M. Frasson, C. Oaida, & S. -H. Yun. (2022). The global surface area variations of lakes and reservoirs as seen from satellite remote sensing. Geophysical Research Letters, vol. 49, e2022GL098987. https://doi.org/10.1029/2022GL098987
Cooley, S.W., J.C. Ryan and L.C. Smith (2021), Human alteration of global water storage variability, Nature, vol. 591, pages 78–81
Das, P., F. Hossain, S. Khan, N. K. Biswas, H. Lee, T. Piman, C. Meechaiya, U. Ghimire, K. Hosen (2022) Reservoir Assessment Tool 2.0: Stakeholder driven Improvements to Satellite Remote Sensing based Reservoir Monitoring, Environmental Modeling and Software, Vol. 157.
Biswas, N. and F. Hossain (2022) A Multi-decadal Analysis of Reservoir Storage Change in Developing Regions, Journal of Hydrometeorology, Vol. 21(1), pp 71-85.
Biswas, N., F. Hossain, M. Bonnema, H. Lee, F. Chishtie (2021). Towards a Global Reservoir Assessment Tool for Predicting Hydrologic Impacts and Operating Patterns of Existing and Planned Reservoirs, Environmental Modeling and Software, Vol. 140.
Zhou, T., Nijssen, B., Gao, H., & Lettenmaier, D. P. (2016). The Contribution of Reservoirs to Global Land Surface Water Storage Variations. Journal of Hydrometeorology, vol 17(1), pp. 309–325. https://doi.org/10.1175/JHM-D-15-0002.1
I urge the authors to explore the above papers very carefully to reframe their introduction and research questions. It will become clear from these papers that just relying on altimeter data on reservoir levels from DAHITI and a corrected Landsat only reservoir area dataset really do not reflect the true state of the art on what we can do today in capturing reservoir dynamics at sub weekly time scales using multiple sensors, wavelengths and innovative methods. The authors will realize that they are answering that question with the most primitive tools and that too by not really defining what the end goal is to be a useful ‘surrogate’ for GHM model development. Monthly reservoir tracking is déjà vu and not hard to do with or without satellites as at that scale, most reservoirs lag the prevailing hydrology (unless they store multiple years of annual runoff). At monthly or longer time scales, GHMs can work reasonably well with parameterizations based on capacity, embedded rule curves and objective functions. However, today we really do not need to do that as we have 40 years of satellite data with the last 10 years of that being very high frequency, multi-sensor to help us track at a granular level how reservoirs have been operated (to help us model them in GHMs).

My request to the authors therefore is to get rid of question 1, tone down or revise substantially section 3.1 and focus most of the paper on section 3.2 where the key contribution lies. Authors should explore using additional datasets -there is now Hydroweb (multiple altimeters), use Sentinel series from 2016 and even MODIS (for larger surface area reservoirs). For the selected reservoirs, the authors have one advantage that cloud cover is minimum. They also do not need to rely on SRTM DEM as that can be very erroneous and over the US, SRTM captured only the ‘free bathymetry’ above the water level that existed during February 2000 when the Shuttle flew. Perhaps authors could test the validity of SRTM DEM with topo maps and published bathymetry and area-elevation curves (consider checking the RESSED database of USGS). The authors use DAHITI only for elevation but as they point out, multiple sensors can be used to generate surface area which apparently DAHITI has done but is not available in the authors’ reservoirs of interest. So why not generate those reservoir areas from multiple sensors and indices by the authors themselves?
There is also the SWOT mission the authors can talk about in a few sentences in the conclusion section (see http://swot.jpl.nasa.gov). The whole premise of SWOT is to generate simultaneously area and elevation so that we don’t have to jury rig the observation system to derive storage change. There is now plenty of ‘help’ resources to help the community build literacy on SWOT (just click on ‘applications’ of the SWOT website). It is also no surprise that DAHITI (elevation) is generating better results for GHMs in section 3.2 and the GRSADs - this is something we keep seeing all the time as all reservoirs experience storage change via elevation change at levels detectable by altimeters. The same can’t be said about surface area changes unless the reservoir is very large and does not have a complex dendritic shape. A lot depends on the shape, shoreline, climate, surrounding terrain of the reservoirs in how well or poorly a specific satellite data will work in tracking surface area or volume change. That is why I reiterate that the authors should focus more on the GHM model validation part rather than the satellite data assessment part in Section 3.1
I noticed many typos that I tried to note down. One is:
Line 144. I don’t think it’s year 286 to 2020 (although we do have many reservoirs built that early still functional in many places)

Citation: https://doi.org/10.5194/hess-2023-215-RC1
- AC1: 'Reply on RC1', Naota Hanasaki, 05 Dec 2023
  
  Our response is shown in the Supplement file.
  
  Citation: https://doi.org/10.5194/hess-2023-215-AC1
RC2:
'Comment on hess-2023-215', Anonymous Referee #2, 16 Oct 2023

The aim of the work of Otta et al. is to use satellite-based storage estimation to improve reservoir representation in Global Hydrological Models. This is an exciting topic because a number of GHMs do not include reservoir storage (Telteu et al. 2022) or generalized reservoir algorithms, which can be quite different from real reservoir operation rules. On-ground observation for most parts of the world is limited for different reasons; therefore, satellite-based data might be a feasible way to bypass the lack of data. The authors tested satellite-based storage estimation against gauge observation and two GHM model results.
The first part – Satellite vs. gauge is already discussed in a more extensive way in other papers e.g. Sorkhabi et al. 2002, Gourgouletis et al. 2022, Hou et al. 2022, and others, either for single reservoirs or on global level. Therefore chapter 3.1 does not add much.
The second part (chapter 3.2.) is the interesting part, but it is a bit small to stand alone. An in deep discussion, why the 2 GHMs representation of monthly and annual average storage can be improved (I think an average 0.5 correlation is far from good representation). Maybe some variation of the reservoir operations (some parameter change) would be nice, especially as the title of the paper is “… to validate reservoir operations.”. The paper has potential , but it needs major changes.
L 40f: Telteu et al. 2022 give a good overview which GHM uses reservoirs and how
L 56f: This lines a repeating L. 51-55. Instead you can cite Zajac et al. 2017 who did a global (real global) analysis on the effect of dams and lakes on streamflow.
L 60f: There are quite a number of studies that evaluate satellite-derived altimetry and the use for reservoir studies. As mentioned before, some of them are more elaborated than this paper on comparing satellite vs. gauge. But the authors are right: “… there remains a need to establish a method... for GHM validation” and I would say the paper shows a need for improvement of reservoir representation in GHM.
L 144: really year 286?
L 145 if the write approx. than 7000 km3 without digits is ok.
L220 Table 1: Maybe add a column for Database GRBD
L224: You can write a bit more about this assumption. There are databases e.g. Hou et al 2022 (GloLakes). You used Hoover dam, there is even a paper looking in detail for this lake mead Li et al. 2022 (Constructing Reservoir Area–Volume–Elevation Curve from TanDEM-X DEM Data)
L 230: Figure 1: This fig looks like fig 3 in Busker et al. 2019 which I think is ok. But you have to explain the dots and a ΔS in the middle would be nice.
L 238: Please explain Sc (I assume maximum storage). Maybe explain here the difficulties to get Ac and Hc from digital elevation models. Here you used the GRBD database which is based on SRTM data?
L 254 later on, you also use GRSA_ISIMIP. Please explain here.
L266: Some of the equations are very basic and fits more into a book of statistics. Equ 7 is not really needed.
L285: Also basic statistics. Maybe interesting if you replace fig 2 with some real examples from your 7 reservoirs. Maybe even 3 reservoirs to show the variability of seasonal variability in your 7 reservoirs example (e.g. Hover,Fort Peck, Wesley)
L 286. It should be Raw (a-d) and … (e-h)
L 290-305: I do not think correlation and NSE has to be explained in this detail. I think that is done 75724 times before. No equations needed.
L 320: Fig 3 After normalization r should stay the same anyway. For f) the comparison to Dahiti does not make much sense. After normalization the NSE for F) becomes worse?
L 326 What is GRSAD_ISIMIP?
L 360: Fig4 What would be interesting to see is the variance of the seasonal variability. Later on in fig 7 we see that the GCMs have a hard time to represent the seasonal variability in some cases. Is the seasonal part very different each year or is it quite stable even in year of drought (e.g. Hoover and Glen from year 2000 onward)?
L415: Figure 5ff: The graphics of S2 is a bit too much spaghetti plot, but still find the supplement figures much more interesting than the average values.
L460f: I would not say that the model simulations are good. The standard for good should be higher than a correlation of 0.5. The authors do not have to make the results better.
L499: As mentioned before there are studies and databases looking at this in detail.
L502: Not currently available. I think there are at least some efforts toward this e.g. GloLakes and others. Referring to a publication from 2012 is not enough.
L 508: I cannot find a detailed discussion of Glen Canyon dam in 3.1.2. But this involves also a discussion on digital elevation products like using SRTM, Aster, Tandem.
Uncertainties: the GHMs uncertainty is missing. This is about satellite vs. Gauge. But the uncertainty of reservoir representation in GHMs is not in yet. But you showed already some e.g. uncertainty of climate forcing . Why is the climate forcing so relevant for Wesley but not for Toledo? What is the uncertainty from reservoir operation in the models (parameter uncertainty and operation uncertainty). Some variation of reservoir control could be done here.
L 533: For 2 out of 7 it is not satisfactory (at least a correlation < 0.5 is not satisfactory for me)
L540: Here it would be nice to answer the question . Why is climate forcing more important for some reservoirs than for others?
L540: Maybe to mention: You cannot compare the effect of forcing by absolute values but only by correlation. Selecting the best meteo forcing by discharge seems still the better method. A multi-objective method would be even better -e.g. snow, evaporation, reservoirs, discharge.
L544: It is not clear if the Dahiti data are general have a low temporal resolution for a majority of global reservoirs or only for your selected 7 reservoirs. If it has a low temporal for a majority of reservoirs this conclusion can be drawn from the very beginning and another satellite product should be used for testing against GRSAD.
L554: It seems that the quality of the elevation model really matters here. There are a number of DGMs e.g. SRTM, Aster, Tandem and hydrological composites of them e.g. Merit.
L 562: For me the paper shows exactly this. The error between satellite and gauge, between different climate forcing (with some exceptions) is quite small compared to the error of GHM vs. satellite or gauge. There is quite a room for improvement in reservoir representation in GHMs. But simply mentioning that refining models, better data and additional factors is way to short for this paper. Why are some reservoirs are very well represented by the models and some not e.g. why is seasonality at Hoover dam so bad and why does it fit so well for Toledo. You might say that was not the purpose of your paper, but chapter 3.1 is not really new, so you have to put something into chapter 3.2.

Citation: https://doi.org/10.5194/hess-2023-215-RC2
- AC2: 'Reply on RC2', Naota Hanasaki, 05 Dec 2023
  
  Our response is shown in the Supplement file.
  
  Citation: https://doi.org/10.5194/hess-2023-215-AC2

Status: closed

RC1:
'Comment on hess-2023-215', Anonymous Referee #1, 12 Sep 2023

This is an interesting study. The authors are exploring the use of satellite remote sensing to validate reservoir operations in global hydrological models (GHMs). I think if we take the perspective of the GHMs and their need to be improved as a modeling framework by moving away from assumptions on reservoir operations and parameterizations (which GHMs have a lot of), then this is a valuable study. However, if this study is just exploring if we can use satellite data to track reservoirs, I don’t think this study is innovative or adds much to the body of knowledge. That topic has been addressed and most of the relevant questions on that that the authors pose in this study have been answered much more vigorously and in more detail over the last 10 years. So the first question posed by the authors is really redundant here (we know the answer): Can satellite-based storage estimation data serve as a surrogate for ground truth data?
The authors have used too simplistic methods of using only altimeter from DAHITI (with limited sampling in space and time) and Landsat (optical) based reservoir area dataset from GRSADs. Today when we have a fleet of satellite sensors (optical, SAR, altimeters) that collectively provide more robust, high frequency and more accurate tracking of reservoirs in terms of surface area or elevation to generate storage change. So I think the authors should reorganize the paper and reduce section 3.1 and focus more on 3.2 as the key focus of the paper where there is a lot to learn for the GHM community trying to improve representation of reservoirs on global models (which also fits perfectly into the theme of the special issue of HESS).
Let me share some key references the authors miss in capturing the state of the art of satellite-based reservoir tracking:
Bonnema, M., C.H. David, R.P. d. M. Frasson, C. Oaida, & S. -H. Yun. (2022). The global surface area variations of lakes and reservoirs as seen from satellite remote sensing. Geophysical Research Letters, vol. 49, e2022GL098987. https://doi.org/10.1029/2022GL098987
Cooley, S.W., J.C. Ryan and L.C. Smith (2021), Human alteration of global water storage variability, Nature, vol. 591, pages 78–81
Das, P., F. Hossain, S. Khan, N. K. Biswas, H. Lee, T. Piman, C. Meechaiya, U. Ghimire, K. Hosen (2022) Reservoir Assessment Tool 2.0: Stakeholder driven Improvements to Satellite Remote Sensing based Reservoir Monitoring, Environmental Modeling and Software, Vol. 157.
Biswas, N. and F. Hossain (2022) A Multi-decadal Analysis of Reservoir Storage Change in Developing Regions, Journal of Hydrometeorology, Vol. 21(1), pp 71-85.
Biswas, N., F. Hossain, M. Bonnema, H. Lee, F. Chishtie (2021). Towards a Global Reservoir Assessment Tool for Predicting Hydrologic Impacts and Operating Patterns of Existing and Planned Reservoirs, Environmental Modeling and Software, Vol. 140.
Zhou, T., Nijssen, B., Gao, H., & Lettenmaier, D. P. (2016). The Contribution of Reservoirs to Global Land Surface Water Storage Variations. Journal of Hydrometeorology, vol 17(1), pp. 309–325. https://doi.org/10.1175/JHM-D-15-0002.1
I urge the authors to explore the above papers very carefully to reframe their introduction and research questions. It will become clear from these papers that just relying on altimeter data on reservoir levels from DAHITI and a corrected Landsat only reservoir area dataset really do not reflect the true state of the art on what we can do today in capturing reservoir dynamics at sub weekly time scales using multiple sensors, wavelengths and innovative methods. The authors will realize that they are answering that question with the most primitive tools and that too by not really defining what the end goal is to be a useful ‘surrogate’ for GHM model development. Monthly reservoir tracking is déjà vu and not hard to do with or without satellites as at that scale, most reservoirs lag the prevailing hydrology (unless they store multiple years of annual runoff). At monthly or longer time scales, GHMs can work reasonably well with parameterizations based on capacity, embedded rule curves and objective functions. However, today we really do not need to do that as we have 40 years of satellite data with the last 10 years of that being very high frequency, multi-sensor to help us track at a granular level how reservoirs have been operated (to help us model them in GHMs).

My request to the authors therefore is to get rid of question 1, tone down or revise substantially section 3.1 and focus most of the paper on section 3.2 where the key contribution lies. Authors should explore using additional datasets -there is now Hydroweb (multiple altimeters), use Sentinel series from 2016 and even MODIS (for larger surface area reservoirs). For the selected reservoirs, the authors have one advantage that cloud cover is minimum. They also do not need to rely on SRTM DEM as that can be very erroneous and over the US, SRTM captured only the ‘free bathymetry’ above the water level that existed during February 2000 when the Shuttle flew. Perhaps authors could test the validity of SRTM DEM with topo maps and published bathymetry and area-elevation curves (consider checking the RESSED database of USGS). The authors use DAHITI only for elevation but as they point out, multiple sensors can be used to generate surface area which apparently DAHITI has done but is not available in the authors’ reservoirs of interest. So why not generate those reservoir areas from multiple sensors and indices by the authors themselves?
There is also the SWOT mission the authors can talk about in a few sentences in the conclusion section (see http://swot.jpl.nasa.gov). The whole premise of SWOT is to generate simultaneously area and elevation so that we don’t have to jury rig the observation system to derive storage change. There is now plenty of ‘help’ resources to help the community build literacy on SWOT (just click on ‘applications’ of the SWOT website). It is also no surprise that DAHITI (elevation) is generating better results for GHMs in section 3.2 and the GRSADs - this is something we keep seeing all the time as all reservoirs experience storage change via elevation change at levels detectable by altimeters. The same can’t be said about surface area changes unless the reservoir is very large and does not have a complex dendritic shape. A lot depends on the shape, shoreline, climate, surrounding terrain of the reservoirs in how well or poorly a specific satellite data will work in tracking surface area or volume change. That is why I reiterate that the authors should focus more on the GHM model validation part rather than the satellite data assessment part in Section 3.1
I noticed many typos that I tried to note down. One is:
Line 144. I don’t think it’s year 286 to 2020 (although we do have many reservoirs built that early still functional in many places)

Citation: https://doi.org/10.5194/hess-2023-215-RC1
- AC1: 'Reply on RC1', Naota Hanasaki, 05 Dec 2023
  
  Our response is shown in the Supplement file.
  
  Citation: https://doi.org/10.5194/hess-2023-215-AC1
RC2:
'Comment on hess-2023-215', Anonymous Referee #2, 16 Oct 2023

The aim of the work of Otta et al. is to use satellite-based storage estimation to improve reservoir representation in Global Hydrological Models. This is an exciting topic because a number of GHMs do not include reservoir storage (Telteu et al. 2022) or generalized reservoir algorithms, which can be quite different from real reservoir operation rules. On-ground observation for most parts of the world is limited for different reasons; therefore, satellite-based data might be a feasible way to bypass the lack of data. The authors tested satellite-based storage estimation against gauge observation and two GHM model results.
The first part – Satellite vs. gauge is already discussed in a more extensive way in other papers e.g. Sorkhabi et al. 2002, Gourgouletis et al. 2022, Hou et al. 2022, and others, either for single reservoirs or on global level. Therefore chapter 3.1 does not add much.
The second part (chapter 3.2.) is the interesting part, but it is a bit small to stand alone. An in deep discussion, why the 2 GHMs representation of monthly and annual average storage can be improved (I think an average 0.5 correlation is far from good representation). Maybe some variation of the reservoir operations (some parameter change) would be nice, especially as the title of the paper is “… to validate reservoir operations.”. The paper has potential , but it needs major changes.
L 40f: Telteu et al. 2022 give a good overview which GHM uses reservoirs and how
L 56f: This lines a repeating L. 51-55. Instead you can cite Zajac et al. 2017 who did a global (real global) analysis on the effect of dams and lakes on streamflow.
L 60f: There are quite a number of studies that evaluate satellite-derived altimetry and the use for reservoir studies. As mentioned before, some of them are more elaborated than this paper on comparing satellite vs. gauge. But the authors are right: “… there remains a need to establish a method... for GHM validation” and I would say the paper shows a need for improvement of reservoir representation in GHM.
L 144: really year 286?
L 145 if the write approx. than 7000 km3 without digits is ok.
L220 Table 1: Maybe add a column for Database GRBD
L224: You can write a bit more about this assumption. There are databases e.g. Hou et al 2022 (GloLakes). You used Hoover dam, there is even a paper looking in detail for this lake mead Li et al. 2022 (Constructing Reservoir Area–Volume–Elevation Curve from TanDEM-X DEM Data)
L 230: Figure 1: This fig looks like fig 3 in Busker et al. 2019 which I think is ok. But you have to explain the dots and a ΔS in the middle would be nice.
L 238: Please explain Sc (I assume maximum storage). Maybe explain here the difficulties to get Ac and Hc from digital elevation models. Here you used the GRBD database which is based on SRTM data?
L 254 later on, you also use GRSA_ISIMIP. Please explain here.
L266: Some of the equations are very basic and fits more into a book of statistics. Equ 7 is not really needed.
L285: Also basic statistics. Maybe interesting if you replace fig 2 with some real examples from your 7 reservoirs. Maybe even 3 reservoirs to show the variability of seasonal variability in your 7 reservoirs example (e.g. Hover,Fort Peck, Wesley)
L 286. It should be Raw (a-d) and … (e-h)
L 290-305: I do not think correlation and NSE has to be explained in this detail. I think that is done 75724 times before. No equations needed.
L 320: Fig 3 After normalization r should stay the same anyway. For f) the comparison to Dahiti does not make much sense. After normalization the NSE for F) becomes worse?
L 326 What is GRSAD_ISIMIP?
L 360: Fig4 What would be interesting to see is the variance of the seasonal variability. Later on in fig 7 we see that the GCMs have a hard time to represent the seasonal variability in some cases. Is the seasonal part very different each year or is it quite stable even in year of drought (e.g. Hoover and Glen from year 2000 onward)?
L415: Figure 5ff: The graphics of S2 is a bit too much spaghetti plot, but still find the supplement figures much more interesting than the average values.
L460f: I would not say that the model simulations are good. The standard for good should be higher than a correlation of 0.5. The authors do not have to make the results better.
L499: As mentioned before there are studies and databases looking at this in detail.
L502: Not currently available. I think there are at least some efforts toward this e.g. GloLakes and others. Referring to a publication from 2012 is not enough.
L 508: I cannot find a detailed discussion of Glen Canyon dam in 3.1.2. But this involves also a discussion on digital elevation products like using SRTM, Aster, Tandem.
Uncertainties: the GHMs uncertainty is missing. This is about satellite vs. Gauge. But the uncertainty of reservoir representation in GHMs is not in yet. But you showed already some e.g. uncertainty of climate forcing . Why is the climate forcing so relevant for Wesley but not for Toledo? What is the uncertainty from reservoir operation in the models (parameter uncertainty and operation uncertainty). Some variation of reservoir control could be done here.
L 533: For 2 out of 7 it is not satisfactory (at least a correlation < 0.5 is not satisfactory for me)
L540: Here it would be nice to answer the question . Why is climate forcing more important for some reservoirs than for others?
L540: Maybe to mention: You cannot compare the effect of forcing by absolute values but only by correlation. Selecting the best meteo forcing by discharge seems still the better method. A multi-objective method would be even better -e.g. snow, evaporation, reservoirs, discharge.
L544: It is not clear if the Dahiti data are general have a low temporal resolution for a majority of global reservoirs or only for your selected 7 reservoirs. If it has a low temporal for a majority of reservoirs this conclusion can be drawn from the very beginning and another satellite product should be used for testing against GRSAD.
L554: It seems that the quality of the elevation model really matters here. There are a number of DGMs e.g. SRTM, Aster, Tandem and hydrological composites of them e.g. Merit.
L 562: For me the paper shows exactly this. The error between satellite and gauge, between different climate forcing (with some exceptions) is quite small compared to the error of GHM vs. satellite or gauge. There is quite a room for improvement in reservoir representation in GHMs. But simply mentioning that refining models, better data and additional factors is way to short for this paper. Why are some reservoirs are very well represented by the models and some not e.g. why is seasonality at Hoover dam so bad and why does it fit so well for Toledo. You might say that was not the purpose of your paper, but chapter 3.1 is not really new, so you have to put something into chapter 3.2.

Citation: https://doi.org/10.5194/hess-2023-215-RC2
- AC2: 'Reply on RC2', Naota Hanasaki, 05 Dec 2023
  
  Our response is shown in the Supplement file.
  
  Citation: https://doi.org/10.5194/hess-2023-215-AC2

Kedar Otta, Hannes Müller Schmied, Simon N. Gosling, and Naota Hanasaki

Supplement

https://doi.org/10.5194/hess-2023-215-supplement

Data sets

Use of satellite remote sensing to validate reservoir operations in global hydrological models: a case study from the CONUS Kedar Otta, Hannes Müller Schmied, Simon N. Gosling, and Naota Hanasaki https://doi.org/10.5281/zenodo.8291850

Kedar Otta, Hannes Müller Schmied, Simon N. Gosling, and Naota Hanasaki

Viewed

Total article views: 1,267 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
912	300	55	1,267	99	58	61

HTML: 912
PDF: 300
XML: 55
Total: 1,267
Supplement: 99
BibTeX: 58
EndNote: 61

Views and downloads (calculated since 12 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	320	76	9	405
Oct 2023	73	20	4	97
Nov 2023	36	11	0	47
Dec 2023	86	20	4	110
Jan 2024	67	13	2	82
Feb 2024	23	9	3	35
Mar 2024	28	16	4	48
Apr 2024	25	9	5	39
May 2024	33	24	6	63
Jun 2024	59	15	3	77
Jul 2024	30	1	5	36
Aug 2024	26	7	0	33
Sep 2024	23	5	0	28
Oct 2024	11	6	0	17
Nov 2024	13	7	1	21
Dec 2024	14	13	0	27
Jan 2025	13	14	4	31
Feb 2025	13	7	2	22
Mar 2025	13	11	3	27
Apr 2025	6	16	0	22

Cumulative views and downloads (calculated since 12 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	320	76	9	405
Oct 2023	73	20	4	97
Nov 2023	36	11	0	47
Dec 2023	86	20	4	110
Jan 2024	67	13	2	82
Feb 2024	23	9	3	35
Mar 2024	28	16	4	48
Apr 2024	25	9	5	39
May 2024	33	24	6	63
Jun 2024	59	15	3	77
Jul 2024	30	1	5	36
Aug 2024	26	7	0	33
Sep 2024	23	5	0	28
Oct 2024	11	6	0	17
Nov 2024	13	7	1	21
Dec 2024	14	13	0	27
Jan 2025	13	14	4	31
Feb 2025	13	7	2	22
Mar 2025	13	11	3	27
Apr 2025	6	16	0	22

Viewed (geographical distribution)

Total article views: 1,178 (including HTML, PDF, and XML) Thereof 1,178 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 20 Apr 2025

Short summary

Reservoirs play important roles in hydrology and water resources management globally and are incorporated into many Global Hydrological Models. Their simulations are, however, poorly validated due to the lack of available long-term in-situ observation data globally. Here we investigated the applicability of the latest satellite-based reservoir storage estimations in the contiguous US. We found that those products are useful for validating reservoir storage simulations when they are normalized.


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