Towards Affordable 3D Physics-Based River Flow Rating: Application Over Luangwa River Basin

Samboko, Hubert T.; Schurer, Sten; Savenije, Hubert H.G.; Makurira, Hodson; Banda, Kawawa; Winsemius, Hessel

doi:https://doi.org/10.5194/gi-2022-21

Preprints

https://doi.org/10.5194/gi-2022-21

Preprints

14 Dec 2022

| 14 Dec 2022

Status: a revised version of this preprint is currently under review for the journal GI.

Towards Affordable 3D Physics-Based River Flow Rating: Application Over Luangwa River Basin

Hubert T. Samboko, Sten Schurer, Hubert H.G. Savenije, Hodson Makurira, Kawawa Banda, and Hessel Winsemius

Abstract. Unmanned aerial vehicles (UAVs), affordable precise Global Navigation Satellite System hardware, echo sounders, open-source 3D hydrodynamic modelling software, and freely available satellite data have opened up opportunities for a robust, affordable, physics-based approach to monitor river flows. In short, the hardware can be used to produce the geometry. 3D hydrodynamic modelling offers a framework to establish relationships between river flow and state variables such as width and depth, while satellite images with surface water detection methods or altimetry records can be used to operationally monitor flows through the established rating curve. Uncertainties in the data acquisition may propagate into uncertainties in the relationships found between discharge and state variables. Variations in acquired geometry emanate from the different ground control point (GCP) densities and distributions which are used during photogrammetry-based terrain reconstruction. In this study, we develop a rating curve using affordable data collection methods and basic principles of physics. The specific objectives were to: determine how the rating curve based on a 3D hydraulic model compares with conventional methods; investigate the impact of geometry uncertainty on estimated discharge when applied in a hydraulic model; and investigate how uncertainties in continuous observations of depth and width from satellite platforms propagate into uncertainties in river flow estimates using the rating curves obtained. The study shows comparable results between the 3D and traditional river rating discharge estimations. The rating curve derived on the basis of 3D hydraulic modelling was within a 95 % confidence interval of the traditional gauging based rating curve. The physics-based estimation requires determination of the roughness coefficient within the permanent bed and the floodplain using field observation as both the end of dry and wet season. Furthermore, the study demonstrates that variations in the density of GCPs beyond an optimal number (9) has no significant influence on the resultant rating relationships. Finally, the study observes that it depends on the magnitude of the flow which state variable approximation (water level & river width) is most promising to use. Combining stage appropriate proxies (water level when the floodplain is entirely filled, and width when the floodplain is filling) in data limited environments yields more accurate discharge estimations. The study was able to successfully apply low cost technologies for accurate river monitoring through hydraulic modelling. In future studies, a larger amount of in-situ gauge readings may be considered so as to optimise the validation process.

Received: 25 Nov 2022 – Discussion started: 14 Dec 2022

Hubert T. Samboko et al.

Status: final response (author comments only)

RC1: 'Comment on gi-2022-21', Anonymous Referee #1, 10 Feb 2023

This paper proposed a method for estimating river flow rate using a 3D flow model and field data. The objective is clear and important. However, the methods and composition of methods are quite common and practical, so a scientific novelty required for an original scientific paper is not enough in the present form.
I am afraid I misunderstand some details, but in my opinion, main scientific and engineering weaknesses of the hypothes of the manuscript are as follows:

1) Use of one rating curve for a channel consisting of a low-flow channel and floodplain (river engineers use two or more rating curves for such a river section better to explain the water level/flow rate relation)

2) Superiority of 3D model compared with 1D model sounds not fair (HEC-RAS had overestimated in Figure 9 for high flood but not sure how the 1D model was calibrated. In the Annex, it was said that the roughness was calibrated for high flow. If so, why was it overestimated in Figure 9? I think a 1D model can account for combined roughness parameters (not confident HEC-RAS can do, but accounting for the combined roughness of a compound channel with 1D flow model is standard.)
Point-to-point (mainly minor) comments:

L15: suggest adding 'multi-beam' before echo sounders (if you want to

include single beam one, use parenthesis)

L17: hardware(s)?

L24: 'determine how .... methods' incomplete sentence?

L25: 'the' hydraulic model?

L29: Meaning of 'physics-based' should be defined in the abstract if use it in the abstract. (Of course, we can guess that author is thinking 3D flow modelling is physics-based, but 1D flow model can be said physics-bases since the shallow-water equations are based on the Navier-Stokes equations which is used for 3D flow model.)

L30: permanent -> stable or immobile (or some others?) ('permanent' sounds something like (very rigid) bedrock but the target site seems sand bar)

L34: 'is most promising to use' A bit vague and logical flow of the sentence is not clear.

L36: remove 'b' before 'hydraulic'

L44: 'implement' would come prior to 'validate'?

L45: flow rate would be one of the most important inputs for the flow model (maybe authors see the flow rate as output, but such standpoint is not explained yet).

L59: 'It is within this technological gap....' hard to read?

L65: 'The process of applying' -> 'Distributing and surveying'?

L81: The sentence seems incomplete?

L91: 'within' ->'for' (not confident)

L92: 'a number of times until' -> 'with different flow rates'?

L99: Combining DEM (obtained by LiDAR or photogrammetrically) and bathymetry (obtained by echo-sounding) are quite common in river engineering, so maybe introducing not only authors' output but other works would be good.

L122, 'the 2 other sites' a bit vague? 'the two sites discussed in previous works (one or two citations)'?

L228: add 'software' before 'D-Flow Flexible...'?

L132: 'in thickness' -> 'their thickness'?

L162: 'point cloud' suddenly appeared.

L164: 'does not affect the water levels' is it realistic? (For subcritical, non-uniform, varied flow, local water level is affected by the downstream flow, I think. Of course, I understand what you want to say, but you can say it with different expressions)

L164: 'A small selection ... is taken' not clear.

Citation: https://doi.org/10.5194/gi-2022-21-RC1
RC2: 'Comment on gi-2022-21', Anonymous Referee #1, 10 Feb 2023

L 186: 'the coordinates of known surface velocities' -> ' the surface velocity distribution'?
L 187: 'the coordinates of known water levels' -> ' the water level profile'?
L 190: better to show the equation of MAD for improve clarity (it's not strong suggestion but maybe help some readers to understand)
L202: 'iterations which estimated the water level based on slope' a bit vague, and better to explain more details. I think we can have results quite similar to the result of WARMA if the 'expert' do the iteration?
L222: Is this a sub-section title?
L232: '=/-' -> '+/\'?
Does the table necessary? (Surface velocity seems to be used for both calibration and validation, it's good if there is some discussion about it)
Table 2: How were the distribution and sample size of three properties discussed in the table? (the result of current metre shows no minimum. Is this show the problem of the current metre survey? (it's hard comment so able to skip))
L261: A bit difficult to understand how this conclusion can be obtained from table 2. (LSPIV shows the minimum with 0.015 s/m^(1/3) but why choose 0.013 as a conclusion)

Citation: https://doi.org/10.5194/gi-2022-21-RC2
RC3:
'Comment on gi-2022-21', Anonymous Referee #1, 10 Feb 2023

L273: remove 'of' at '100 m3/s of were'?
L274: 'four rating curves derived from D4DFM; one based on....' Misleading? (two rating curves were based on D4DFM but two others were not?)
L299: Spell out 'OLS'?
L299: P_{bias} and E_{ns}: compared with 17GCPs result or WARMA? (maybe with 17GCPs but better to indicate(
L314: Better to indicate the reference if the uncertainty used here (also for L329)
L323: 'more stable roughness coefficient' a bit unclear.
L326: 'schematized' and 'schematization' are used very close, maybe can be rephrase to improve readability.
L342: figure(s)?
Bibliography of Kim. Y (2006) could be edited.
Figure 9: Labels 'measurements HEC-RAS, D3DFM, Combined roughness' are a bit confusing. something like'estimate with HEC-RAS, D3DFM (single roughness), D4DFM (combined roughness)'?
Figure 10: The regression line seems weighted to high flow. Is there any reason? (based on annex B, it seems logs are applied to both axes)
Figure 13: It's just my interest, but it's better to add the data points of 1200 and 1400 m3/s, and I want to see if there are some disconnections in the relation, like Figure 12.
That's all.

Citation: https://doi.org/10.5194/gi-2022-21-RC3
- AC1:
  'Reply on RC3', Hubert Samboko, 13 Feb 2023
  Dear Referee #1
  Firstly, on behalf of the authors of this manuscript allow me to express my utmost gratitude for taking the time to read through our manuscript so thoroughly. The comments you have provided are clearly articulated. We are confident that implementation of the suggested recommendations/corrections will improve the quality of our work significantly.
  We acknowledge that the novelty of this paper may not have been well articulated and there is therefore room for improvement. In our point of view the manuscript is novel in that it suggests merging the dry bathymetry with the wet bathymetry through linear integration specifically for the purposes of application in a hydraulic model using low-cost technologies. The authors were successful in generating this seamless bathymetry with a high level of accuracy through application accessible advanced technologies (low-cost RTK GNSS, low cost UAV, open source software: Cloud Compare/Python). Some of these technologies (particularly the RTK GNSS) have previously been out of reach for some water resource authorities due to financial limitations as exemplified in the poorly gauged Luangwa Basin. We aim to utilise advancement in these technologies to as remotely as possible monitor rivers in ungauged/poorly gauged basins.
  Below we attempt to respond to the 2 main areas that have been pointed out by the reviewer:
  Use of one rating curve for a channel consisting of a low-flow channel and floodplain (river engineers use two or more rating curves for such a river section better to explain the water level/flow rate relation)
  
  This recommendation is strongly related to the element of novelty which has been identified by the reviewer. In our geometry-based approach, we generate a continuous relationship between the conveyance and the water level. This is new. One of the impediments to the application of a single rating curve is the difficulty in extracting accurate low flow bathymetric information. The integration of an affordable yet accurate RTK GNSS on the ADCP allows for extraction of similarly accurate bathymetric data throughout the terrain.
  We hope to make this clear by adding a detailed description within the introduction of why we choose to apply one all-encompassing rating curve.
  Superiority of 3D model compared with 1D model sounds not fair
  
  The authors acknowledge that the text may seem biased towards the 3 D model. The authors propose that the 3D model is not so much more accurate, rather, more robust in the information that it provides. For instance, with 3D models we are able to model lateral movement of water. Furthermore, when modelling specific areas of interest (e.g. flood prone area) it is useful to visualise the flow of water even after the banks of the river are fully submerged.
  We propose to restructure the text to make it clear that the aim is not to discredit the capability of a 1 D model, rather, we attempt to maximise the capabilities and advantages of the high resolution which in characteristic of a 3D model.
  
  Finally, we appreciate the comments that the reviewer has titled ‘Point-to-point (mainly minor) comments:’
  We have gone through each point and will be making corrections to each comment.
  
  Citation: https://doi.org/10.5194/gi-2022-21-AC1
  - RC4: 'Reply on AC1', Anonymous Referee #1, 20 Feb 2023
    
    > In our geometry-based approach, we generate a continuous relationship between the conveyance and the water level. This is new. One of the impediments to the application of a single rating curve is the difficulty in extracting accurate low flow bathymetric information.
    I understand what you want to say, and also afraid 're-inventing wheel' thing (especially the last sentence is not new, in essence), so I suggest carefully checking and distinguishing which part is new, and which part is something revisiting thing. I think it's ok to view the same thing from a different viewpoint (for this paper, using an economic approach) but better to describe comparing with existing findings.
    
    Citation: https://doi.org/10.5194/gi-2022-21-RC4

RC5: 'Comment on gi-2022-21', Anonymous Referee #2, 24 Feb 2023

The study focuses on the use of cutting-edge technology to estimate river discharge. Although the methods and results are interesting, the research fails to clearly explain the contribution to science. The advantage of using this UAV-system vs traditional methods is not clear. These traditional methods are cheaper and therefore more affordable compared with the described systems (i.e., drone, GNS, eco sounder, etc). However, traditional methods are based on point measurements and heterogeneity of the river geometry is not contemplated. In contrast it seems that the UAV-system can capture the spatial variability of the channel shape, hence probably a more accurate flow discharge.

I would recommend refocussing the problem statement showing scientific evidence of how point measurements fail to estimate river discharge. There is a hint of this during the introduction (e.g., L68 and L74) but is not addressed very clearly. For instance this should be emphasised in the abstract - perhaps it can be the opening sentence. Also, in the introduction, it would be good to show studies on the implication of different spatial resolution for the estimation of river discharge (use of Lidar and satellite data for river geometry, and gauging stations). It will also be interesting to show the spatial resolution of your UAV-system. How does this compare with the available DEM data (e.g., Lidar, satellite, cm vs m)?

In general, there is a lack of discussion in the document. For example, you should discuss that comparing your method against traditional ones provides some sort of validation, and in fact as mentioned, more measures are needed for the reliability of your method. Also, it would be important to address in the discussion the differences between using a 1D model vs a 3D model in your results. Additionally, I believe that more than one rating curve should be used. Regardless of this, you need to discuss the implications of only using one rating curve.

I consider this research has substantial scientific merits. However, you need to improve the structure and the writing to make it more fluent and easier to read. I encourage to use the below suggestions and comments for an improvement. If the suggestions are not appropriate to the scope of the research, please reconsider better wording to transmit the idea.

Please find below specific comments.

L17 – What do you mean with “hardware”? Would it not be better to use the word “system” Also, the sentence “In short, the hardware can be used to produce the geometry” is confusing. Do you mean river geometry? The sentence, as it is, seeming to be incomplete or somehow needs to be related to the previous sentence or the following one.

L22- I recommend mentioning the novelty/contribution in the abstract. This can be place before objectives and after briefly explaining the problem.

L24 – Instead of using semicolon, I would recommend alphabetic numerating of the objectives (a, b, c, etc). This will allow the reader to easily differentiate between them.

L32 –Using the number 9 in parentheses is confusing, it only makes sense when reading the methods in the paper. I recommend removing it and leaving the sentence "beyond an optimal number" or change it to "beyond an optimal threshold of 9 GCPs".

L36 – remove “d”.

L44 – I would use the word “estimation” rather than “monitoring”. Monitoring can be confused with sensing, then it is better to clarify that models are useful tools for prediction rather than monitoring/sensing.

L45 – Use the word “apply” rather than “implement”.

L75 – I would remove the word "robust". It could be argued that more measurements and rating curves are needed to make the method robust. I leave it for your consideration.

L80 – Research questions are objectives rewritten as questions. Although, there is nothing wrong with this, it is repeated information. If the authors want to leave the research questions, I suggest modifying them. I leave it for your consideration.

L86 to L95 – It is not easy to follow the steps as they are written. I suggest numerating them (i, ii, etc).

L119 - Were the measurements of flow and water level contemporary with those of GCP and bathymetry? If so, what year was it (2022)? Could you please add in a table the data collection date, or maybe add this in table 1.

L210 – Add name of variables (O = observation, P, x, etc)

L222 – Seems incomplete.

L237 – Results should be section “3”. Previous section is “2 Material and Methods”.

L312 – I don’t recall you mentioning satellite data in the “2 Material and Methods” section. You need to add this in section “2.7”?

L353 -Where is the “discussion” section? I consider it very important to discuss the differences of using a 3D model vs the 1D model. Also, the use of more than one rating curve (if using only one discuss why?). The advantages of your method over others, etc

Figure 2 and Figure 5 - Legend and scales need to be bigger. It is difficult to read.

Figure 3 – I don’t think this figure provides important information. I would remove it also because it does not follow the same format as other figures.

Figure 4. – Subfigures require identification (a & b) and a respective legend. Also, they look identical to me. Make evident the “volumised and cut on both sides”. What do the colours mean? -depth (m)? Add a colour bar.

Figure 5 – I think figure 4 and figure 5 can be a single one (a, b, and c).

Figure 6- Use identification for subfigures (a, b, c, etc) and add their respective legend (5 GCP, 9 GCP, etc).

Figure 7 – Bigger legend.

Figure 8 – This is a good example of a figure.

In general, in the text there is a discrepancy in the format. Sometimes you use a space between new lines (L135, L160, L180, etc), sometimes you don't (e.g., L50, L59, L114, etc). Also, there are tabs where they shouldn't be (L221 and L214). In terms of the figures, you use different colour and letter sizes (e.g., Figure 1 a, b labels in black vs figure 2 a, b labels in parentheses and in orange).

Citation: https://doi.org/10.5194/gi-2022-21-RC5

AC2: 'Reply on RC5', Hubert Samboko, 21 Mar 2023

Firstly, we would like to sincerely thank the reviewer for the meticulous attention to detail especially in regards to the papers potential contribution to science. The reviewer provides useful comments and recommendations which we believe will significantly improve the manuscript if implemented adequately. We acknowledge the shortcomings that have been identified in terms of comparison between the UAV system and the traditional estimation methods. It is particularly true that the UAV system comes across as more expensive. However, we suggest that these tools applied in the UAV system (Drone, low-cost RTK GNSS, eco sounder etc.) are in fact more affordable than previously available physics based technologies (Lidar, differential GNSS, ADCP etc.). In hindsight, Instead of suggesting that the UAV system is cheaper than the traditional one, we could consider saying, the more physics based system in particular has become considerably more affordable. This lower cost in combination with benefits of a higher spatial resolution makes it is worthwhile to implement for more accurate non-contact flow estimation.

As suggested, we will refocus the problem statement showing scientific evidence of how point measurements fail to estimate river discharge. This will also be added to the abstract as well as the Introduction in the form of literature review of the impact of spatial resolution on river discharge estimation accuracy.

In general the reviewer points us in the right direction with respect to the need to add more discussion of results (1D vs 3D model, reliability of results, number of rating curves required). Finally, we have gone through all 24 specific comments and will be making amendments to the output as advised.

Line	Comment	Action
1	L17 -What do you mean with “hardware”? Would it not be better to use the word “system” Also, the sentence “In short, the hardware can be used to produce the geometry” is confusing. Do you mean river geometry? The sentence, as it is, seeming to be incomplete or somehow needs to be related to the previous sentence or the following one.	Replacing the word hardware with system. We indeed are refereeing to the ‘river geometry’
2	L22- I recommend mentioning the novelty/contribution in the abstract. This can be place before objectives and after briefly explaining the problem.	We will add the novelty which will refer to the potential of utilising advanced technologies with a higher spatial resolution to estimate flow.
3	L24 – Instead of using semicolon, I would recommend alphabetic numerating of the objectives (a, b, c, etc). This will allow the reader to easily differentiate between them.	We will add the alphabetic numbering to objectives.
4	L32 –Using the number 9 in parentheses is confusing, it only makes sense when reading the methods in the paper. I recommend removing it and leaving the sentence "beyond an optimal number" or change it to "beyond an optimal threshold of 9 GCPs".	Change as suggested to ‘beyond an optimal threshold of 9 GCPs’
5	L36 – remove “d”.	Change as suggested
6	L44 – I would use the word “estimation” rather than “monitoring”. Monitoring can be confused with sensing, then it is better to clarify that models are useful tools for prediction rather than monitoring/sensing.	Change as suggested
7	L45 – Use the word “apply” rather than “implement”.	Change as suggested
8	L75 – I would remove the word "robust". It could be argued that more measurements and rating curves are needed to make the method robust. I leave it for your consideration.	Change as suggested. Indeed more rating curves wold be needed to prove that the method is robust
9	L80 – Research questions are objectives rewritten as questions. Although, there is nothing wrong with this, it is repeated information. If the authors want to leave the research questions, I suggest modifying them. I leave it for your consideration.	Change as suggested. We will opt to rephrase the questions as suggested
10	L86 to L95 – It is not easy to follow the steps as they are written. I suggest numerating them (i, ii, etc).	Change as suggested
11	L119 - Were the measurements of flow and water level contemporary with those of GCP and bathymetry? If so, what year was it (2022)? Could you please add in a table the data collection date, or maybe add this in table 1.	The WARMA measurements are not contemporary. They were recorded from 1948 to 2002. However, at the time of measuring the GCP and bathymetry, flow measurements were also made. We will add the data collection to table 1
12	L210 – Add name of variables (O = observation, P, x, etc)	Change as suggested
13	L222 – Seems incomplete.	Heading error to be corrected. The sentence should be a title
14	L237 – Results should be section “3”. Previous section is “2 Material and Methods”.	Title correction to be made
15	L312 – I don’t recall you mentioning satellite data in the “2 Material and Methods” section. You need to add this in section “2.7”?	Will add satellite data to 2.7 as suggested
16	L353 -Where is the “discussion” section? I consider it very important to discuss the differences of using a 3D model vs the 1D model. Also, the use of more than one rating curve (if using only one discuss why?). The advantages of your method over others, etc	We will add a discussion section on these topics
17	Figure 2 and Figure 5 - Legend and scales need to be bigger. It is difficult to read.	Change as suggested
18	Figure 3 – I don’t think this figure provides important information. I would remove it also because it does not follow the same format as other figures.	Remove as suggested
19	Figure 4. – Subfigures require identification (a & b) and a respective legend. Also, they look identical to me. Make evident the “volumised and cut on both sides”. What do the colours mean? -depth (m)? Add a colour bar.	Edit as suggested
20	Figure 5 – I think figure 4 and figure 5 can be a single one (a, b, and c).	Change as suggested
21	Figure 6- Use identification for subfigures (a, b, c, etc) and add their respective legend (5 GCP, 9 GCP, etc).	Change as suggested
22	Figure 7 – Bigger legend.	Change as suggested
23	Figure 8 – This is a good example of a figure.	Noted, Thank you. Will use as reference
24	In general, in the text there is a discrepancy in the format. Sometimes you use a space between new lines (L135, L160, L180, etc), sometimes you don't (e.g., L50, L59, L114, etc). Also, there are tabs where they shouldn't be (L221 and L214). In terms of the figures, you use different colour and letter sizes (e.g., Figure 1 a, b labels in black vs figure 2 a, b labels in parentheses and in orange).	Will adjust all these discrepancies as advised.

Citation: https://doi.org/10.5194/gi-2022-21-AC2

Hubert T. Samboko et al.

Data sets

Python Scripts Hubert Samboko https://doi.org/10.4121/21557148

Hubert T. Samboko et al.

Viewed

Total article views: 974 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
859	88	27	974	11	11

HTML: 859
PDF: 88
XML: 27
Total: 974
BibTeX: 11
EndNote: 11

Views and downloads (calculated since 14 Dec 2022)

Month	HTML	PDF	XML	Total
Dec 2022	118	22	7	147
Jan 2023	137	9	0	146
Feb 2023	210	26	12	248
Mar 2023	162	13	4	179
Apr 2023	127	9	0	136
May 2023	102	9	4	115
Jun 2023	3	0	3

Cumulative views and downloads (calculated since 14 Dec 2022)

Month	HTML	PDF	XML	Total
Dec 2022	118	22	7	147
Jan 2023	137	9	0	146
Feb 2023	210	26	12	248
Mar 2023	162	13	4	179
Apr 2023	127	9	0	136
May 2023	102	9	4	115
Jun 2023	3	0	3

Viewed (geographical distribution)

Total article views: 940 (including HTML, PDF, and XML) Thereof 940 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 02 Jun 2023

Short summary

The study investigates how low cost technology can be applied in data scarce catchments to improve water resource management. More specifically, we investigate how drone technology can be combined with low-cost RTK GNSS equipment and subsequently applied in a 3 D hydraulic model so as to generate more physically based rating curves.


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