the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Sep 2022
Deriving transmission losses in ephemeral rivers using satellite imagery and machine learning
Antoine Di Ciacca
Scott Wilson
Jasmine Kang
Thomas Wöhling
Abstract. Transmission losses are the loss in the flow volume of a river as water moves downstream. These losses provide crucial ecosystem services, particularly in ephemeral and intermittent river systems. Transmission losses can be quantified at many scales using different measurement techniques. One of the most common methods is differential gauging of river flow at two locations. An alternative method for non-perennial rivers is to replace the downstream gauging location by visual assessments of the wetted river length on satellite images. We used this approach to estimate the transmission losses in the Selwyn River (Canterbury, New Zealand) using 147 satellite images collected between March 2020 and May 2021. The location of the river drying front was verified in the field on five occasions and seven differential gauging campaigns were conducted to ground-truth the losses estimated from the satellite images. The transmission loss point data obtained using the wetted river lengths and differential gauging campaigns were used to train an ensemble of random forest models to reconstruct the hourly time series of transmission losses and their uncertainties. Our results show that the Selwyn river transmission losses ranged between 0.25 and 0.65 m3/s/km during most of the 1-year study period. However, shortly after a flood peak the losses could reach up to 1.5 m3/s/km. These results enabled us to improve our understanding of the Selwyn River groundwater – surface water interactions and provide valuable data to support water management. We argue that our framework can easily be adapted to other ephemeral rivers and to longer time series.
-
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
(5010 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(5010 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Antoine Di Ciacca et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-833', Howard Wheater, 20 Sep 2022
Overall this paper makes a useful contribution to an important issue – the quantification of channel transmission losses, which is often a dominant process in ephemeral streams. The combination of remotely sensed and in situ data provides a useful database, and points to the more general applicability of remote sensing data in this context.
The paper would benefit from clarification of several points of detail, as noted below. Also some more incisive discussion of aspects of the observed data would strengthen the paper, and associated discussion of the potential physical causes of variability in the results over different periods (as discussed in the comparison with other studies). Recommendations fir further work would be useful.
The abstract does not indicate how transmission losses can be derived from a wetted river length. An additional sentence is needed for clarification.
Line 18 typo McMahon (also line 50)
l46 downstream of
l94/95 similar to
l106 flows for
l110 ‘inland plains’ is unclear. Is coastal plains what is meant, or is there some differentiation intended between an inland plain and a coastal plain? If so, some explanation is needed.
Fig 1 typo ‘losing’
l143 derivative
l148 discussion about time before peak unclear at this point
l152 similar to that adopted….
l153 for clarity, recall what Walters did – i.e. use the volume gauged upstream to estimate the lost discharge
l171 explain what is the difference between the river bed and the active river channel – how are these identified?
l181 the linear model fitted in App C masks some interesting aspects of the data, which need discussion. For example there are segments that show both strong gains at some times and strong losses at others. What are possible explanations and how do these effects reflect on the very simple assumption of a linear model? We need some process insights here.
l186 what is meant by the transmission loss time series? Given the spatial complexity and the multiple measurements, this phrase is ambiguous without further explanation.
l215 The ‘estimated transmission losses vary in time’ is unclear. They also vary in space as well as time, so some clarification is needed.
l224 need to explain where the peak flow that is referred to occurred – presumably this is at the permanent gauging station (clearly a) peak flow is very different when downstream points near the wetting front are considered, and b) there is transmission time for peak flow to propagate downstream)
l227 ‘transmission losses were maximum’ is unclear. I assume that what is meant is ‘transmission losses estimated using the modelled relationship with flow at the gauging station’. Above, this was described as estimated, but not here?
l232 fig 4 caption needs some qualification. These are estimated transmission losses based on the stage at the gauged hydrograph location. (similar comment for Fig 6 caption)
Fig C1 shows some sections (below 750m) change from losing to gaining – so complex surface water groundwater interactions
l301 differ in
l314 how could hydrological variability be expected to affect the results? Presumably this relates to groundwater effects? Some thought/discussion is needed, perhaps linking to the observed variability in response shown in App C.
l395 – concluding comments – some thought should be given as to how to further develop insights into the response of this system. It seems to be crying out for some basic monitoring of groundwater. Is there really no data and no monitoring planned?
l411 developed
Citation: https://doi.org/10.5194/egusphere-2022-833-RC1 - AC1: 'Reply on RC1', Antoine Di Ciacca, 21 Oct 2022
-
RC2: 'Comment on egusphere-2022-833', Anonymous Referee #2, 29 Sep 2022
Dear Authors,
The paper presents a framework to estimate transmission losses from satellite imagery and predict their hourly time series using random forest regressors. The study is interesting and shows novelty, but it must be improved in many aspects in my point of view. Moreover, there are some critical points that should be carefully addressed by the Authors. Please, find below my comments and suggestions.
Major comments:
- Page 2, Line 42-44: “Furthermore, as noted by Cook (2015), it is unusual for two gauging stations to be located on the same river without intervening tributaries. Therefore, quantifying transmission losses from two existing gauging stations is rarely possible.” I disagree. In reality, for large dryland rivers, where the most studies on channel transmission losses were undertaken, upstream river discharge is much larger than runoff produced between the streamgauges. Considering allogeneic dryland rivers, this runoff is practically null. Therefore, quantifying transmission losses from two or more existing gauging stations is perfectly possible in drylands.
- The described perceptual model of the surface-groundwater interaction of the study site (2 Study site) must be much better spatially presented, showing profiles along and across the main river and aquifer units.
- As you described in the study site, is it correct to say that the water lost in the ephemeral losing reach is immediately available again downstream?
- The ephemeral reach is always a losing one? Even during high floods?
- Page 6, Line 137: Why did you use five inflection points for the rating curve?
- 3.2.1 Transmission losses derived from the river drying front locations: the main problem of this methodology is there are just five days of comparison between the GPS and satellite drying front positions, although daily satellite images were available. Therefore, more fieldwork should be done, in order to properly estimate the uncertainty on the satellite wetted river length estimation.
- Page 7, Line 180: you wrote that “the higher uncertainties are typically associated with shallow and low flow in the smaller braids.” However, your fitted linear model showed a rather different result with higher uncertainties related to larger flows.
- Page 13, Line 245: “... and the effect of the peaks becomes an important control” How?
- 4.3 Reconstructed transmission loss time series: What did we learn about the transmission losses in the Selwyn River when the machine learning approach was applied? If there is nothing to add to our understanding of the process, I suggest either excluding it or to use another time series model.
- You should compare your study with previous studies conducted on other ephemeral streams, including those from other climates. It is fundamental to place your findings in the context of transmission loss research.
Minor comments:
- I suggest moving the Figures A1, B1 and C1 from Appendix to the main text.
- Page 7, Lines 190-192: “In the course of the model development, more predictors (e.g. river flow, water temperature, groundwater level, date) have been tested but they appeared to not improve significantly the predictions.” Have you tried any statistical criterion, such as AIC?
- Please, reconsider the terminology of “reconstructed” transmission losses, because reconstruction of time series is a quite different topic. You should use just “predicted” transmission losses.
- Page 12, Lines 235-236: “The estimated transmission losses range from 0.14 to 1.5m3/s/km. Most of the estimated losses (56%) are below 0.6m3/s/km and correspond mainly to baseflow periods and river drying phases.” Please, provide a box-plot of the transmission losses, and add more statistical details.
- Conclusions: It is not necessary to use citations in the conclusion.
Citation: https://doi.org/10.5194/egusphere-2022-833-RC2 - AC2: 'Reply on RC2', Antoine Di Ciacca, 21 Oct 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-833', Howard Wheater, 20 Sep 2022
Overall this paper makes a useful contribution to an important issue – the quantification of channel transmission losses, which is often a dominant process in ephemeral streams. The combination of remotely sensed and in situ data provides a useful database, and points to the more general applicability of remote sensing data in this context.
The paper would benefit from clarification of several points of detail, as noted below. Also some more incisive discussion of aspects of the observed data would strengthen the paper, and associated discussion of the potential physical causes of variability in the results over different periods (as discussed in the comparison with other studies). Recommendations fir further work would be useful.
The abstract does not indicate how transmission losses can be derived from a wetted river length. An additional sentence is needed for clarification.
Line 18 typo McMahon (also line 50)
l46 downstream of
l94/95 similar to
l106 flows for
l110 ‘inland plains’ is unclear. Is coastal plains what is meant, or is there some differentiation intended between an inland plain and a coastal plain? If so, some explanation is needed.
Fig 1 typo ‘losing’
l143 derivative
l148 discussion about time before peak unclear at this point
l152 similar to that adopted….
l153 for clarity, recall what Walters did – i.e. use the volume gauged upstream to estimate the lost discharge
l171 explain what is the difference between the river bed and the active river channel – how are these identified?
l181 the linear model fitted in App C masks some interesting aspects of the data, which need discussion. For example there are segments that show both strong gains at some times and strong losses at others. What are possible explanations and how do these effects reflect on the very simple assumption of a linear model? We need some process insights here.
l186 what is meant by the transmission loss time series? Given the spatial complexity and the multiple measurements, this phrase is ambiguous without further explanation.
l215 The ‘estimated transmission losses vary in time’ is unclear. They also vary in space as well as time, so some clarification is needed.
l224 need to explain where the peak flow that is referred to occurred – presumably this is at the permanent gauging station (clearly a) peak flow is very different when downstream points near the wetting front are considered, and b) there is transmission time for peak flow to propagate downstream)
l227 ‘transmission losses were maximum’ is unclear. I assume that what is meant is ‘transmission losses estimated using the modelled relationship with flow at the gauging station’. Above, this was described as estimated, but not here?
l232 fig 4 caption needs some qualification. These are estimated transmission losses based on the stage at the gauged hydrograph location. (similar comment for Fig 6 caption)
Fig C1 shows some sections (below 750m) change from losing to gaining – so complex surface water groundwater interactions
l301 differ in
l314 how could hydrological variability be expected to affect the results? Presumably this relates to groundwater effects? Some thought/discussion is needed, perhaps linking to the observed variability in response shown in App C.
l395 – concluding comments – some thought should be given as to how to further develop insights into the response of this system. It seems to be crying out for some basic monitoring of groundwater. Is there really no data and no monitoring planned?
l411 developed
Citation: https://doi.org/10.5194/egusphere-2022-833-RC1 - AC1: 'Reply on RC1', Antoine Di Ciacca, 21 Oct 2022
-
RC2: 'Comment on egusphere-2022-833', Anonymous Referee #2, 29 Sep 2022
Dear Authors,
The paper presents a framework to estimate transmission losses from satellite imagery and predict their hourly time series using random forest regressors. The study is interesting and shows novelty, but it must be improved in many aspects in my point of view. Moreover, there are some critical points that should be carefully addressed by the Authors. Please, find below my comments and suggestions.
Major comments:
- Page 2, Line 42-44: “Furthermore, as noted by Cook (2015), it is unusual for two gauging stations to be located on the same river without intervening tributaries. Therefore, quantifying transmission losses from two existing gauging stations is rarely possible.” I disagree. In reality, for large dryland rivers, where the most studies on channel transmission losses were undertaken, upstream river discharge is much larger than runoff produced between the streamgauges. Considering allogeneic dryland rivers, this runoff is practically null. Therefore, quantifying transmission losses from two or more existing gauging stations is perfectly possible in drylands.
- The described perceptual model of the surface-groundwater interaction of the study site (2 Study site) must be much better spatially presented, showing profiles along and across the main river and aquifer units.
- As you described in the study site, is it correct to say that the water lost in the ephemeral losing reach is immediately available again downstream?
- The ephemeral reach is always a losing one? Even during high floods?
- Page 6, Line 137: Why did you use five inflection points for the rating curve?
- 3.2.1 Transmission losses derived from the river drying front locations: the main problem of this methodology is there are just five days of comparison between the GPS and satellite drying front positions, although daily satellite images were available. Therefore, more fieldwork should be done, in order to properly estimate the uncertainty on the satellite wetted river length estimation.
- Page 7, Line 180: you wrote that “the higher uncertainties are typically associated with shallow and low flow in the smaller braids.” However, your fitted linear model showed a rather different result with higher uncertainties related to larger flows.
- Page 13, Line 245: “... and the effect of the peaks becomes an important control” How?
- 4.3 Reconstructed transmission loss time series: What did we learn about the transmission losses in the Selwyn River when the machine learning approach was applied? If there is nothing to add to our understanding of the process, I suggest either excluding it or to use another time series model.
- You should compare your study with previous studies conducted on other ephemeral streams, including those from other climates. It is fundamental to place your findings in the context of transmission loss research.
Minor comments:
- I suggest moving the Figures A1, B1 and C1 from Appendix to the main text.
- Page 7, Lines 190-192: “In the course of the model development, more predictors (e.g. river flow, water temperature, groundwater level, date) have been tested but they appeared to not improve significantly the predictions.” Have you tried any statistical criterion, such as AIC?
- Please, reconsider the terminology of “reconstructed” transmission losses, because reconstruction of time series is a quite different topic. You should use just “predicted” transmission losses.
- Page 12, Lines 235-236: “The estimated transmission losses range from 0.14 to 1.5m3/s/km. Most of the estimated losses (56%) are below 0.6m3/s/km and correspond mainly to baseflow periods and river drying phases.” Please, provide a box-plot of the transmission losses, and add more statistical details.
- Conclusions: It is not necessary to use citations in the conclusion.
Citation: https://doi.org/10.5194/egusphere-2022-833-RC2 - AC2: 'Reply on RC2', Antoine Di Ciacca, 21 Oct 2022
Peer review completion
Journal article(s) based on this preprint
Antoine Di Ciacca et al.
Antoine Di Ciacca et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 296 | 80 | 13 | 389 | 2 | 2 |
- HTML: 296
- PDF: 80
- XML: 13
- Total: 389
- BibTeX: 2
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(5010 KB) - Metadata XML