Characterization of Cold Land Hydrological Processes by Integrating In-Situ Snowpack Observations with a Land Surface Model in the Yellowstone River Basin, USA

Kang, Do Hyuk

doi:https://doi.org/10.5194/egusphere-2023-2776

Preprints

https://doi.org/10.5194/egusphere-2023-2776

Preprints

30 Nov 2023

| 30 Nov 2023

Characterization of Cold Land Hydrological Processes by Integrating In-Situ Snowpack Observations with a Land Surface Model in the Yellowstone River Basin, USA

Do Hyuk Kang

Abstract. In the eastern region of the North American Continental Divide in the upper Colorado Rockies, this study demonstrates that enhancing streamflow predictability from May to July in the Yellowstone River Basin is enabled. This streamflow improvement is achieved by employing a land surface hydrology model in the watershed, coupled with an updated winter precipitation weather forcing dataset. Utilizing 13 snowpack telemetry stations from the US Department of Agriculture in the Yellowstone River Basin, the paper calculates ratios between a baseline simulated snowpack from the initial land surface model application and the observed snowpack. The average ratio serves as a constant multiplier for the existing snowfall weather forcing applied in the second land surface model simulation. As a result of the second simulation, the streamflow predictability reaches a Nash-Sutcliffe Efficiency (NSE) of 0.91, in contrast to the baseline simulation's 0.73 NSE during peak streamflow periods. The study also explores cold land hydrological processes, particularly those related to snowmelt-driven streamflow. In addition to streamflow, two land surface variables such as snowpack and soil moisture are assessed against in-situ snowpack and satellite-based soil moisture observations in the Yellowstone River Basin. The comparisons reveal that the peak of soil moisture is mainly driven by springtime snowmelt and diminishes in the summer. The findings are confirmed by both land surface model simulations and satellite-borne soil moisture observations. Another noteworthy discovery is that soil infiltration properties in the Yellowstone River Basin are wetter than the western Continental Divide in North America, resulting in amplified streamflow in the eastern side despite similar levels of snowmelt runoff on either side of the Continental Divide in the upper Colorado Rockies in the United States.

Received: 21 Nov 2023 – Discussion started: 30 Nov 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

Do Hyuk Kang

Status: closed

RC1:
'Comment on egusphere-2023-2776', Anonymous Referee #1, 05 Jan 2024

Dear editor,
In the manuscript “Characterization of Cold Land Hydrological Processes by Integrating In-Situ Snowpack Observations with a Land Surface Model in the Yellowstone River Basin, USA”, the author improves simulated streamflow in the Yellowstone river by adjusting the snowfall inputs based on the difference in observed and simulated snow water equivalents. Additionally, the author attributes the streamflow disparity between the Yellowstone and Snake rivers to the difference in soil infiltration in the respective areas. Although the methods used in the manuscript are sound, I would recommend the manuscript to be rejected as it is poorly written, contains unsubstantiated conclusions and carries little significance for the scientific community. Below is a more expansive description of my main arguments as well as a list of line-by-line comments.

Scientific significance
The main body of the manuscript consists of the efforts to improve simulated streamflow in the upper and lower Yellowstone river by adjusting ERA-5 snowfall inputs. As mentioned in the manuscript (lines 226-227), ERA-5 can suffer from snow undercatch (i.e. insufficient snowfall) due to the data’s reliance on weather station observations that are often biased towards valleys in mountainous areas. Therefore, the author adjusts the ERA-5 snowfall with a constant factor based on the average difference between the simulated snow water equivalent under a unadjusted ERA-5 simulation and several observations.
However, this approach only confirms what is already known, namely that ERA-5 likely estimates insufficient snowfall in the region. Moreover, the limitations and future use of the approach are not explored. For example, although streamflow simulations are a bit better under the adjusted snowfall amounts (figure 6), the simulated snow water equivalent actually gets worse for several stations (figure 2). As the limitations and usefulness of the approach are not discussed, the manuscript only reflects that streamflow simulations (with a specific hydrological model setup) with more snowfall perform better in the Yellowstone river basin than streamflow simulations without.
Unsubstantiated conclusions
There are several issues with the results presented in section 3, that subsequently inform the conclusions in section 4, that I would like to mention. Specifically, they relate to (1) the comparisons between differently calibrated simulations, (2) the comparison between observed and simulated soil moisture and (3) the conclusions drawn from the Yellow river and Snake river simulation differences.
Although the manuscript attributes the streamflow improvements solely to the increase in snowfall, the snowfall adjustment is not the only difference between the simulations. Rather, the manuscript mentions (lines 315-317) that a re-calibration took place between these simulations as well. The impact of this recalibration on the hydrological states and fluxes is never discussed. Therefore it may well be that the streamflow improvement cannot only be attributed to the snowfall adjustment.
The manuscript also compares the simulated soil moisture with the SMAP satellite observations. However, there a two substantial limitations to this comparison. Firstly, the manuscript only compares simulations and observations on two individual points. I see no reason why a spatial comparison would not be possible and more informative (single points do not accurately reflect model performance). Secondly, the manuscript compares the SMAP signal to the average soil moisture content in either the whole soil column or the third soil layer. However, SMAP only measures the soil moisture content in the upper 5 centimeters of the soil (which is only a part of the first soil layer). Therefore this comparison cannot be valid.
The manuscript also compares the streamflow differences between the Yellow and Snake rivers east and west of the North American Continental Divide. The manuscript concludes that this difference is driven by the higher infiltration rates in the Snake river basin. However, higher infiltration does not directly result in lower streamflow. In fact, the routing scheme used traditionally includes the baseflow, which is essentially infiltrated soil moisture. Therefore, infiltrated water is not lost and cannot be the reason for the streamflow difference.
Writing
Both the article structure as well as its language could be substantially improved. The structure and language improvements are provided in the line-by-line comments below, but some examples are: there is no discussion, the key knowledge gap (and its references) are located in the results, the approach to adjust the snowfall is introduced during a data description (and before the model or meteorology description), figures contain errors, code availability reference is incorrect and the comparison between the Yellowstone and Snake rivers has no relation to the rest of the manuscript.

Line-by-line comments
Line 8 “enhancing streamflow predictability (…) is enabled”: How can predictability be enabled?
Line 9-10 “updated winter precipitation weather forcing dataset”: “corrected snowfall”
Line 13-14 “streamflow predictability”: “streamflow performance”
Line 14: Metrics are given during the calibration period, not the validation period!
Line 19-22: why these additional results? They seem unconnected to the rest of the manuscript (different region, different research question).
Line 19-20 “ soil infiltration properties (…) are wetter”: How can infiltration properties be wetter?
Line 14 “snowpack embraces significant importance”: Embraces?
Lines 25-40: Strong focus on climate change (and wildfires?) even though the article does not focus on climate change.
Lines 48-49: “with various researchers establishing this body of knowledge”: Would remove
Lines 55-56: “However, the application of land surface hydrology models to the Yellowstone river (…) is relatively recent”: This is not the case, there are many studies that use land surface or hydrological models in this region. An example from the VIC model used in the manuscript is the 1997 paper by Bart Nijssen (Streamflow simulation for continental-scale river basins; DOI 10.1029/96WR03517), where he actually simulates the whole world.
Line 58 “the understanding of hydrological processes remains limited”: Use the word “understanding” with care, as model simulations also do not directly contribute to “understanding”. Rather, they use the current understanding to predict trends/changes.
Section 1.1: Generally too much detail about the region and the streamflow gauges, but no map.
Line 106 “fo”: “for”
Line 139-141: Why do you assume the meteorology is wrong and not the model?
Lines 155-156: “The VIC model is configured with 18 rows and 29 columns”: This provides no information to the reader
Lines 162-163 “determined using the minimum and maximum elevations in the watershed”: Determined how?
Line 168 “ROUT”: Newly introduced but never mentioned or explained.
Lines 191-193: Repeat of information.
Line 199 “In cases of deep (above 200mm SWE) and shallow (below 200mm SWE)”: So in all cases.
Section 3: Generally too much new information, which should have been given in the introduction or methods, in the results. Moreover, there is a strong need for quantification of the results (KGE, correlation, mean squared error and more).
Line 209 “Following the methodology of the prior study conducted by Kang and Jung (2023)”: What methodology?
Line 225 “European Reanalysis Assimilation 5 (ERA-5)”: “ECMWF Reanalysis v5 (ERA5)”!
Line 258-259 “both SWE and soil moisture simulations align well with in-situ SWE and satellite-based soil moisture observations”: I do not agree, performance is rather poor.
Line 321: “ VICMF (0.86) (…) is outperformed by VICBL (0.58)”: Does a higher NSE not indicate that VICMF it performs better?
Line 326-327 “is more pronounced at the outlet than in the upstream”: results show exactly the opposite, with a larger NSE improvement in the upstream than at the outlet.
Table 3 and 4: Why are performance metrics selected during the calibration period? This should be during the validation period.
Lines 331-332: “specifically, in 2002, 2008 and 2018”: Why select these years?
Figure 6: three timeseries where there should be two. Middle is duplicate of top.
Lines 386-391: Why not directly compare the soil parameters in the input?
Lines 408-409: This is not a conclusion.
Lines 436-437: This GitHub repository only contains the model code, not the “code used to analyze the data and generate the figures in this study”!
Sections headers are incorrect (twice section 3 and 3.2)
No discussion!

Citation: https://doi.org/10.5194/egusphere-2023-2776-RC1
- AC1: 'Reply on RC1', Dohyuk Kang, 29 Jan 2024
  
  Dear reviewer,
  Please find 'reply to comments'. I will upload other materials including 1) a track-changed manuscript and 2) a clean manuscript after receiving the 2nd review.
  Thank you very much.
  
  Best,
  DK.
  
  Citation: https://doi.org/10.5194/egusphere-2023-2776-AC1
RC2: 'Comment on egusphere-2023-2776', Anonymous Referee #2, 18 Feb 2024

Dear Editor,
This paper mainly discussed the streamflow predictability by simply applying an adjusted factor to the winter precipitation, and the discussion of the impact of snowmelt on streamflow and soil moisture is not very new knowledge in snowmelt dominated regions. Therefore, I think the findings and scope of this paper lack enough novelty to be published to the Cryosphere.
I read through RC1's comments and RC1 shares many similar comments as mine, here I mainly add a few more.
1. the multiplier is a constant in the paper, but the underestimation can have spatial heterogeneity and temporal variation, which is not considered in this paper.
2. both VIC_MF and VIC_BL are recalibrated, making it hard to say the change is mainly due to the adjustment of snow fall.
3. line 31-40, focus on wildfires, need a better linkage to the focus of the paper, which is about streamflow rather than drought/wildfire.
4. line 321: replace but with and.
5. line 323: subscript MF.

Citation: https://doi.org/10.5194/egusphere-2023-2776-RC2

Status: closed

RC1:
'Comment on egusphere-2023-2776', Anonymous Referee #1, 05 Jan 2024

Dear editor,
In the manuscript “Characterization of Cold Land Hydrological Processes by Integrating In-Situ Snowpack Observations with a Land Surface Model in the Yellowstone River Basin, USA”, the author improves simulated streamflow in the Yellowstone river by adjusting the snowfall inputs based on the difference in observed and simulated snow water equivalents. Additionally, the author attributes the streamflow disparity between the Yellowstone and Snake rivers to the difference in soil infiltration in the respective areas. Although the methods used in the manuscript are sound, I would recommend the manuscript to be rejected as it is poorly written, contains unsubstantiated conclusions and carries little significance for the scientific community. Below is a more expansive description of my main arguments as well as a list of line-by-line comments.

Scientific significance
The main body of the manuscript consists of the efforts to improve simulated streamflow in the upper and lower Yellowstone river by adjusting ERA-5 snowfall inputs. As mentioned in the manuscript (lines 226-227), ERA-5 can suffer from snow undercatch (i.e. insufficient snowfall) due to the data’s reliance on weather station observations that are often biased towards valleys in mountainous areas. Therefore, the author adjusts the ERA-5 snowfall with a constant factor based on the average difference between the simulated snow water equivalent under a unadjusted ERA-5 simulation and several observations.
However, this approach only confirms what is already known, namely that ERA-5 likely estimates insufficient snowfall in the region. Moreover, the limitations and future use of the approach are not explored. For example, although streamflow simulations are a bit better under the adjusted snowfall amounts (figure 6), the simulated snow water equivalent actually gets worse for several stations (figure 2). As the limitations and usefulness of the approach are not discussed, the manuscript only reflects that streamflow simulations (with a specific hydrological model setup) with more snowfall perform better in the Yellowstone river basin than streamflow simulations without.
Unsubstantiated conclusions
There are several issues with the results presented in section 3, that subsequently inform the conclusions in section 4, that I would like to mention. Specifically, they relate to (1) the comparisons between differently calibrated simulations, (2) the comparison between observed and simulated soil moisture and (3) the conclusions drawn from the Yellow river and Snake river simulation differences.
Although the manuscript attributes the streamflow improvements solely to the increase in snowfall, the snowfall adjustment is not the only difference between the simulations. Rather, the manuscript mentions (lines 315-317) that a re-calibration took place between these simulations as well. The impact of this recalibration on the hydrological states and fluxes is never discussed. Therefore it may well be that the streamflow improvement cannot only be attributed to the snowfall adjustment.
The manuscript also compares the simulated soil moisture with the SMAP satellite observations. However, there a two substantial limitations to this comparison. Firstly, the manuscript only compares simulations and observations on two individual points. I see no reason why a spatial comparison would not be possible and more informative (single points do not accurately reflect model performance). Secondly, the manuscript compares the SMAP signal to the average soil moisture content in either the whole soil column or the third soil layer. However, SMAP only measures the soil moisture content in the upper 5 centimeters of the soil (which is only a part of the first soil layer). Therefore this comparison cannot be valid.
The manuscript also compares the streamflow differences between the Yellow and Snake rivers east and west of the North American Continental Divide. The manuscript concludes that this difference is driven by the higher infiltration rates in the Snake river basin. However, higher infiltration does not directly result in lower streamflow. In fact, the routing scheme used traditionally includes the baseflow, which is essentially infiltrated soil moisture. Therefore, infiltrated water is not lost and cannot be the reason for the streamflow difference.
Writing
Both the article structure as well as its language could be substantially improved. The structure and language improvements are provided in the line-by-line comments below, but some examples are: there is no discussion, the key knowledge gap (and its references) are located in the results, the approach to adjust the snowfall is introduced during a data description (and before the model or meteorology description), figures contain errors, code availability reference is incorrect and the comparison between the Yellowstone and Snake rivers has no relation to the rest of the manuscript.

Line-by-line comments
Line 8 “enhancing streamflow predictability (…) is enabled”: How can predictability be enabled?
Line 9-10 “updated winter precipitation weather forcing dataset”: “corrected snowfall”
Line 13-14 “streamflow predictability”: “streamflow performance”
Line 14: Metrics are given during the calibration period, not the validation period!
Line 19-22: why these additional results? They seem unconnected to the rest of the manuscript (different region, different research question).
Line 19-20 “ soil infiltration properties (…) are wetter”: How can infiltration properties be wetter?
Line 14 “snowpack embraces significant importance”: Embraces?
Lines 25-40: Strong focus on climate change (and wildfires?) even though the article does not focus on climate change.
Lines 48-49: “with various researchers establishing this body of knowledge”: Would remove
Lines 55-56: “However, the application of land surface hydrology models to the Yellowstone river (…) is relatively recent”: This is not the case, there are many studies that use land surface or hydrological models in this region. An example from the VIC model used in the manuscript is the 1997 paper by Bart Nijssen (Streamflow simulation for continental-scale river basins; DOI 10.1029/96WR03517), where he actually simulates the whole world.
Line 58 “the understanding of hydrological processes remains limited”: Use the word “understanding” with care, as model simulations also do not directly contribute to “understanding”. Rather, they use the current understanding to predict trends/changes.
Section 1.1: Generally too much detail about the region and the streamflow gauges, but no map.
Line 106 “fo”: “for”
Line 139-141: Why do you assume the meteorology is wrong and not the model?
Lines 155-156: “The VIC model is configured with 18 rows and 29 columns”: This provides no information to the reader
Lines 162-163 “determined using the minimum and maximum elevations in the watershed”: Determined how?
Line 168 “ROUT”: Newly introduced but never mentioned or explained.
Lines 191-193: Repeat of information.
Line 199 “In cases of deep (above 200mm SWE) and shallow (below 200mm SWE)”: So in all cases.
Section 3: Generally too much new information, which should have been given in the introduction or methods, in the results. Moreover, there is a strong need for quantification of the results (KGE, correlation, mean squared error and more).
Line 209 “Following the methodology of the prior study conducted by Kang and Jung (2023)”: What methodology?
Line 225 “European Reanalysis Assimilation 5 (ERA-5)”: “ECMWF Reanalysis v5 (ERA5)”!
Line 258-259 “both SWE and soil moisture simulations align well with in-situ SWE and satellite-based soil moisture observations”: I do not agree, performance is rather poor.
Line 321: “ VICMF (0.86) (…) is outperformed by VICBL (0.58)”: Does a higher NSE not indicate that VICMF it performs better?
Line 326-327 “is more pronounced at the outlet than in the upstream”: results show exactly the opposite, with a larger NSE improvement in the upstream than at the outlet.
Table 3 and 4: Why are performance metrics selected during the calibration period? This should be during the validation period.
Lines 331-332: “specifically, in 2002, 2008 and 2018”: Why select these years?
Figure 6: three timeseries where there should be two. Middle is duplicate of top.
Lines 386-391: Why not directly compare the soil parameters in the input?
Lines 408-409: This is not a conclusion.
Lines 436-437: This GitHub repository only contains the model code, not the “code used to analyze the data and generate the figures in this study”!
Sections headers are incorrect (twice section 3 and 3.2)
No discussion!

Citation: https://doi.org/10.5194/egusphere-2023-2776-RC1
- AC1: 'Reply on RC1', Dohyuk Kang, 29 Jan 2024
  
  Dear reviewer,
  Please find 'reply to comments'. I will upload other materials including 1) a track-changed manuscript and 2) a clean manuscript after receiving the 2nd review.
  Thank you very much.
  
  Best,
  DK.
  
  Citation: https://doi.org/10.5194/egusphere-2023-2776-AC1
RC2: 'Comment on egusphere-2023-2776', Anonymous Referee #2, 18 Feb 2024

Dear Editor,
This paper mainly discussed the streamflow predictability by simply applying an adjusted factor to the winter precipitation, and the discussion of the impact of snowmelt on streamflow and soil moisture is not very new knowledge in snowmelt dominated regions. Therefore, I think the findings and scope of this paper lack enough novelty to be published to the Cryosphere.
I read through RC1's comments and RC1 shares many similar comments as mine, here I mainly add a few more.
1. the multiplier is a constant in the paper, but the underestimation can have spatial heterogeneity and temporal variation, which is not considered in this paper.
2. both VIC_MF and VIC_BL are recalibrated, making it hard to say the change is mainly due to the adjustment of snow fall.
3. line 31-40, focus on wildfires, need a better linkage to the focus of the paper, which is about streamflow rather than drought/wildfire.
4. line 321: replace but with and.
5. line 323: subscript MF.

Citation: https://doi.org/10.5194/egusphere-2023-2776-RC2

Do Hyuk Kang

Viewed

Total article views: 193 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
123	52	18	193	10	11

HTML: 123
PDF: 52
XML: 18
Total: 193
BibTeX: 10
EndNote: 11

Views and downloads (calculated since 30 Nov 2023)

Month	HTML	PDF	XML	Total
Dec 2023	27	7	2	36
Jan 2024	42	9	6	57
Feb 2024	37	17	7	61
Mar 2024	7	6	1	14
Apr 2024	8	11	2	21
May 2024	2	2	0	4

Cumulative views and downloads (calculated since 30 Nov 2023)

Month	HTML	PDF	XML	Total
Dec 2023	27	7	2	36
Jan 2024	42	9	6	57
Feb 2024	37	17	7	61
Mar 2024	7	6	1	14
Apr 2024	8	11	2	21
May 2024	2	2	0	4

Viewed (geographical distribution)

Total article views: 189 (including HTML, PDF, and XML) Thereof 189 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 08 May 2024

Short summary

In the upper Colorado Rockies on the eastern side of the North American Continental Divide, this study finds that we can enhance the predictability of river flow in the Yellowstone River Basin from May to July. The enhancement in streamflow is achieved by using a land surface hydrology model that simulates how water moves through the land and combining it with revised winter precipitation weather forcing.


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