Evaluating low-cost topographic surveys for computations of conveyance

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

doi:https://doi.org/10.5194/gi-11-1-2022

Articles | Volume 11, issue 1

https://doi.org/10.5194/gi-11-1-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gi-11-1-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 1

Research article

|

25 Jan 2022

Research article |

| 25 Jan 2022

Evaluating low-cost topographic surveys for computations of conveyance

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

Download

Final revised paper (published on 25 Jan 2022)
Preprint (discussion started on 07 Jul 2021)

Interactive discussion

Status: closed

RC1: 'Comment on gi-2021-22', Anonymous Referee #1, 17 Aug 2021

The present manuscript focuses on the accuracy of UAV photogrammetry products for river geometry reconstructions. This is a crucial point for the implementation of hydraulic models for practical engineering purposes. The work is based on a dataset of images acquired from UAV in the North Eastern part of Zambia. I found the article too weak in several points because of the lack of a complete overview of the literature and of the essential details in the descriptions of methodologies adopted. I think that the paper needs a big improvement before being reconsidered for publication. In the following, I list some major and specific comments that should be addressed carefully.

SPECIFIC COMMENTS

1) Title: The title provided is too generic. I suggest to change it focusing on the UAS technologies for surveying.

2) Abstract. The abstract is too long. It should be a very brief summary of your paper.

3) Section 1. The introduction is too generic and doesn’t focus on the main questions of the paper. Some references are missed or included only in the following sections. Some main research questions (e.g. the impact of lens distortion on geometry accuracy) are introduced only in the final sections. Authors should include a complete overview of previous research on this topic (at line 108-111) in order to evidence the added value of their outcomes.

4) Section 2.

- Please move the research questions at the end of section 1 “Introduction”.

- I suggest to create a new section for “Study site”. To this regard, an overview of the morphological and hydraulic characteristics of the river reach can be useful.

- Respect to the third research question (line 128), please clarify that the objectives are referred to the error estimation of some variables useful for the indirect estimation of the discharge (hydraulic conveyance and slope).

5) Sub-section 2.2.1. Line 145. All the configurations used for the study it should be specified at this point.

6) Section 2.2.4. Line 185: The information relative to computer performance can be useful if a comparison between the two software on computation time is achieved.

7) Section 2.3.2. Line 230. I suggest to specify how the GCP points are spatially distributed along the river pattern especially respect to the vertical variability.

8) Sub-section 3.2: The configuration Brown-Conrady is not described in the main text.

9) Section 3. A separate discussion section should be added in the main structure of the manuscript. This section should include a comparison with other research studies and should be extended to other aspect that play a role in this analysis, i.e. the flight mission planning and the camera settings.

10) Section 3.3. Line 325. It is not clear how the slope is calculated based on photogrammetry products.

11) Section 3.3. Line 336. This step require more details for a better explanation of the procedure.

12) Conclusion and recommendations. Line 377. Please clarify this point in the section “Methods and material”: the number of points used for reconstructions and those for validation.

13) Figures:

- Please improve the overall quality of the figures.

- Generally, captions are not very descriptive. Please modify accordingly.

- Some figures are not described in the main text (e.g. Figure 11, Figure 5b).

- In some figure, useful information is missing: the name of cross-sections (figure 7), the measure units (Figure 11), the flow direction.

Citation: https://doi.org/10.5194/gi-2021-22-RC1

AC1: 'Reply on RC1', Hubert Samboko, 26 Aug 2021

We find these comments extremely helpful and appreciate the time taken to give recommendations. We are in the process of making all relevant adjustment to the manuscript and will revert soon.

Kind Regards
H.S

Citation: https://doi.org/10.5194/gi-2021-22-AC1

AC3: 'Reply on RC1', Hubert Samboko, 20 Sep 2021

We acknowledge and appreciate the comments provided by RC1. We find that the comments were accurate in pointing out weaknesses of the manuscript especially with respect to missing descriptions and literature review of key methodologies/outputs. We therefore kindly submit the following responses to each comment made in order to improve the document.

No	Comments from Reviewers	Authors Response	Authors Changes
1	Title: The title provided is too generic. I suggest to change it focusing on the UAS technologies for surveying.	We note and agree with the comment. The study indeed relies on the results of the UAV products and RTK GNSS (i.e. low cost technologies)	We suggest changing the title to ‘EVALUATING LOW COST TOPOGRAPHIC SURVEYS FOR HYDRAULIC RATING’
2	Abstract. The abstract is too long. It should be a very brief summary of your paper.	We note the lengthiness of the abstract. We identify that we include a long methodology explanation which may not be necessary for an abstract.	Abstract will be shortened. This will be achieved by excluding the elaborate methodology explanation
3.1	The introduction is too generic and doesn’t focus on the main questions of the paper. Some references are missed or included only in the following sections.	We identify the shortcomings of the Introduction. Mainly through the specific exclusion of : 1. Information on the RTK GNSS equipment (Accuracy, cost, method) 2. The gridding or merging approach of the bathymetry 3. The doming effect and lens distortion (brown Conrady, Fixed camera parameter FCP etc.) 4. The influence/importance of the slope	We propose to significantly alter the introduction to add all the important missing information. We will alter the introduction such that the novel attributes of the manuscript as well as the objectives are fully described.
3.2	Some main research questions (e.g. the impact of lens distortion on geometry accuracy) are introduced only in the final sections.	This is a critical missing component.	We will add the missing information earlier on in the manuscript. E.g. the impact of lens distortion accuracy the observation of slope and the merging of dry and wet river profile
3.3	Authors should include a complete overview of previous research on this topic (at line 108-111) in order to evidence the added value of their outcomes.	It is noted that there are references necessary to support the claim that ‘limited studies have investigated how critical factors (GCP number and distribution, lens distortion, slope, free and open source software) can be adjusted to improve hydraulic modelling’	We will add all the missing reference literature on how previous researchers have not focused on 1. How factors can be adjusted to improve elevation models 2. how to combine the wet and dry bathymetry
4.1	Please move the research questions at the end of section 1 “Introduction”.	Noted	The research questions will be moved from section 2 to the end of section 1 as advised.
4.2	I suggest to create a new section for “Study site”. To this regard, an overview of the morphological and hydraulic characteristics of the river reach can be useful.	To give the reader some insight into the environment within which the study is conducted, we agree with the comment and add a ‘study site’ section to describe the river characteristics	We will add a section titled ‘Morphological and hydraulic characteristics of the river reach’
4.3	Respect to the third research question (line 128), please clarify that the objectives are referred to the error estimation of some variables useful for the indirect estimation of the discharge (hydraulic conveyance and slope).	The objective was indeed misleading to reader as it may have been misconstrued as estimation of discharge rather than an estimation of proxy variables such a conveyance and slope	Objective changed to be more specific ‘What impact does utilising elevation models, reconstructed based on different GCP numbers have on conveyance and hydraulic slope’
5	Sub-section 2.2.1. Line 145. All the configurations used for the study it should be specified at this point.	The various configurations were indeed missing. Especially those to do with the GCP numbers and distributions	We will added a paragraph which describes the configurations for the various experiments. Particularly the GCP distribution and density. To aid with configuration visualization, we will add an image of the flight pattern which was used to collect the images.
6	Section 2.2.4. Line 185: The information relative to computer performance can be useful if a comparison between the two software on computation time is achieved.	To be able to authoritatively compare software it is indeed noted that all computer hardware variables need to be mentioned.	We will add more information on the computer performance. In addition we will add a reference to the minimum hardware requirements according to the software developers
7	Section 2.3.2. Line 230. I suggest to specify how the GCP points are spatially distributed along the river pattern especially respect to the vertical variability.	Seeing as literature suggests that GCP distribution is critical to achieve good accuracy, it is noted that it would be useful to explain the distribution methodology used.	We will describe how the GCP are distributes (E.g. the 2-1-2 or checkerboard formation) and also outline how the maximum and minimum elevations were taken into consideration in order to not only have a representative horizontal distribution, but a vertical distribution as well.
8	Sub-section 3.2: The configuration Brown-Conrady is not described in the main text.	Brown Conrady is linked to one of the key factors which determine if photogrammetric geometry will be accurate. Its omission in the main text is an error	We will add a literature review on the Brown Conrady as well as other calibration models to the main text (Introduction). We will also describe some of the configuration which can deal with the doming effect such as FCP in the main text.
9	Section 3. A separate discussion section should be added in the main structure of the manuscript. This section should include a comparison with other research studies and should be extended to other aspect that play a role in this analysis, i.e. the flight mission planning and the camera settings.	There are indeed some factors which are not looked at in this topic but, however need to be mentioned.	We will add a section which takes a closer look at how other studies have assessed the impact of other factors which are not the subject of this particular manuscript such as flight mission planning, camera setting , flight height, speed, direction, light conditions etc.
10	Section 3.3. Line 325. It is not clear how the slope is calculated based on photogrammetry products.	An explanation of how slope is derived is missing.	We will add a description of the method of slope calculation. This will include a brief explanation of a python module called ‘rasterio’ which is able to interpret raster images, and therefore extract elevation values (Z) which correspond to the RTK line (the ‘true’) slope coordinates (X,Y)
11	Section 3.3. Line 336. This step require more details for a better explanation of the procedure.	A more detailed explanation is necessary to aid the reader.	We will break down the procedure into the different processes (extraction, merging, volumising). We will then describe each process individually. E.g. the extraction entails overlaying the wet bathymetry on the DEM and cutting out the shape using a special tool in cloud compare’
12	Conclusion and recommendations. Line 377. Please clarify this point in the section “Methods and material”: the number of points used for reconstructions and those for validation.	A description of the exact number of points used was indeed missing. This is in terms of the GCP vs check-points	We will add a description of all the various GCP configurations which were used to 2.2.1 ‘Flight Plan’
13	Figures: - Please improve the overall quality of the figures.	Noted	We will either adjust or redo some of the figures in terms of the actual image quality, labelling and caption
13.1	Generally, captions are not very descriptive. Please modify accordingly.	Noted	We will adjust all captions such that they fully describe what can be seen in each image
13.2	Some figures are not described in the main text (e.g. Figure 11, Figure 5b).	Noted	We will make sure that all al images are fully described. Figure 5b in particular will be explained is a flow chart related to SfM and MVS. A step by step description will be added to section 2.2.4 ‘processing dry and wet bathymetry’
13.3	In some figure, useful information is missing: the name of cross-sections (figure 7), the measure units (Figure 11), flow direction.	Noted	We will adjust these and other images appropriately

Citation: https://doi.org/10.5194/gi-2021-22-AC3

RC2: 'Comment on gi-2021-22', Anonymous Referee #2, 19 Aug 2021

This paper aims to investigate how well low-cost UAV and RTK-GNSS software and systems can perform for developing topographic surfaces in river environments, potentially for hydraulic modeling purposes. This is an important topic, as UAVs are being increasingly used to collect high-resolution topography, and as pointed out in the paper, their low cost has the potential to greatly benefit managers in developing countries. I believe there are good data presented here, and the results will be of interest to many; however, the paper’s organization and presentation could be improved to more clearly describe what was done, why it was done, and what the important take away points are. I have some general comments and questions that, if addressed, may help make the paper more effective in presenting the most important findings.

My main comment is that it seems the paper needs to be reframed around the apparently novel contributions of the paper, which are not addressed in the current version of the introduction. Many of the most interesting and potentially important results presented in the paper are not related to the objectives set up in the paper’s Introduction. The paper is framed as a study focused on how remote sensing data from UAVs can be optimized, particularly with regard to the number and spatial distribution of ground control points. However, the topographic data collected with the RTK system is as critical or perhaps even more critical for the analyses they present than the UAV data and SfM processing. Indeed, one of the main findings in the paper is that there are diminishing returns as additional GCPs are included in the SfM processing – this is a good outcome to report and readers will want to know that, but many of the results rely substantially on the dense RTK-GNSS bathymetry collected from the canoe, and/or the RTK-GNSS profile of the water edge collected with the rolling cart. So much of the data which are important for the analyses are not collected with UAVs, but the introduction is almost entirely focused on UAVs.

It seems to me that if the paper intends to focus on low-cost but high-resolution characterization of river channel geometry, the novel contributions of this paper would include: 1) the low-cost RTK-GNSS system itself – how accurate is it, how is it put together, how much does it actually cost? (this is not really discussed in the paper). 2) The commercial vs. open source UAV software comparison (this is done in the paper). 3) Assessment of how many and what distribution of GCPs is best for UAV SfM data (this is also done in the paper). 4) The gridding approach for the RTK bathymetry data (discussed but not emphasized in the paper). 5) How to deal with the ‘doming’ effect in SfM datasets (discussed almost entirely in the Results section, but since this is a focus there should be some background on the issue in the introduction).

Additional comments (by line number):

18: clarify in the abstract that GNSS data are used to characterize the subaqueous bathymetry, and UAVs are only used to map the dry surfaces.

165: The description of the UAV flight path is not clear. Was the UAV flown in one direction back and forth (“lawnmower” style) or in two direction back and forth (“checkerboard” style)?

190: The flowchart in Fig 5 needs to be described better. For instance, what is “MVS”?

252: clarify how the slope was extracted – was a plane fit to the DEM? Is the slope computed from the average of dry points?

312: I believe this is the first mention of “Fixed Camera Parameter”. This needs to be described earlier and in more detail. The method is partly described later (line 345) but that is out of place.

354: Figure 13 makes it appear there is a lateral slope of the water surface. Was there? Can we be sure the RTK system is working properly?

479: Figures B4 and B5 do not have legends.

Citation: https://doi.org/10.5194/gi-2021-22-RC2

AC2: 'Reply on RC2', Hubert Samboko, 26 Aug 2021

In adition to comments made by RC1, we find these comments extremely useful and appreciate the time taken to give recommendations. We note the overlap with comments made by RC1, reaffirming the need to make some important changes. We are in the process of making all relevant adjustment to the manuscript and will revert soon.

Kind Regards
H.S

Citation: https://doi.org/10.5194/gi-2021-22-AC2

AC4: 'Reply on RC2', Hubert Samboko, 20 Sep 2021

We acknowledge and appreciate the comments made by RC2. We are greatly encouraged to hear that the reviewer finds the data collected potentially interesting and of practical use to water managers in developing countries. As advised by RC2, we acknowledge the need to reorganise the paper to focus more on the novel attributes such as the ‘gridding’ approach, low-cost GNSS based bathymetry, RTK line. We therefore submit the following point by point responses to the reviewer comments.

	Comments from Reviewers	Authors Response	Authors Changes
1	clarify in the abstract that GNSS data are used to characterize the subaqueous bathymetry, and UAVs are only used to map the dry surfaces.	The manuscript indeed misses out on the opportunity to describe much more about the low-cost GNSS. Which we believe could be revolutionary in terms of access to accurate measurements for researchers with smaller budgets.	We will adjust the abstract to clarify that the GNSS is in fact the key tool for the wet bathymetry reconstruction. We will also add a description of the system and its costs to the main text.
2	The description of the UAV flight path is not clear. Was the UAV flown in one direction back and forth (“lawnmower” style) or in two direction back and forth (“checkerboard” style)?	Given that the flight path is important as mechanism that can be manipulated to reduce the doming effect, it is noted that the specific flight method must be clarified.	We will add a description of the flight path as well as a figure to aid with visualisation.
3	The flowchart in Fig 5 needs to be described better. For instance, what is “MVS”?	The flow-chart which describes the SfM processing of the dry bathymetry was indeed not described. Including terms such as Multi view Stereo (MVS)	We will redo the image to include fully described terms rather than acronyms and we will detail a step by step description of the flowchart in section 2.2.4 ‘processing dry and wet bathymetry’
4	clarify how the slope was extracted – was a plane fit to the DEM? Is the slope computed from the average of dry points?	An explanation of how slope is derived is missing. NB* The slope will be compare to the slope of the RTK line (collected using the rolling cart)	We will add a description of the method of slope calculation. This will include a brief explanation of a python module called ‘rasterio’ which is able to interpret raster images, and therefore extract elevation values (Z) which correspond to the RTK line (the ‘true’) slope coordinates (X,Y).
5	I believe this is the first mention of “Fixed Camera Parameter”. This needs to be described earlier and in more detail. The method is partly described later (line 345) but that is out of place.	The term FCP is indeed misplaced and is supposed to be described in the introduction in-line with methods which can potentially decrease the impact of lens distortion (doming)	We will add a description of FCP to the main text, including references and why it is potentially useful.
6	Figure 13 makes it appear there is a lateral slope of the water surface. Was there? Can we be sure the RTK system is working properly?	We had the opportunity to do simple pre-experimental tests on the accuracy of the RTK system and its accuracy was working. There was no lateral slope of the water surface. However the extreme left bank of the river was inaccessible due to overgrown vegetation. This implies that a small section which would equalise the water levels on the left and right bank is missing.	We will provide an explanation for the apparent lateral slope. It may also be worthwhile to suggest selection of a site which is easily accessible on both sides of the river in future studies. Unfortunately selecting a location which would satisfy this condition and many other conditions such as a straight reach, accessible flood plain, etc. was not straightforward
7	Figures B4 and B5 do not have legends.	Noted	We will add legends to both figure sets.

Citation: https://doi.org/10.5194/gi-2021-22-AC4

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Hubert Samboko on behalf of the Authors (04 Oct 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (11 Oct 2021) by Antonio Annis

RR by Anonymous Reviewer #2 (27 Oct 2021)

Suggestions for revision or reasons for rejection

The authors have made a significant revision of this manuscript, and overall it is much improved from the original submission. The organization of the paper has been especially improved, and shifting the focus to include more on the low-cost RTK-GNSS system helps broaden the impact of this work. My comments, by line number, are below:

1: consider changing the title - seems like it would make more sense to mention channel conveyance rather than hydraulic rating. Perhaps ‘Evaluation of low-cost topographic surveys for computations of river conveyance’?

22: clarify what ‘these’ is referring to

24: ‘on hydraulics’ should just be ‘hydraulics’

48: ‘measurement’ should be ‘measurements’

51: unclear what is meant by ‘time validity of the measurements’

57: I think ‘measurements’ should be ‘calculations’ – you are discussing doing a calculation (not measurement) of discharge here.

95: ‘there are more factors to consider’ – for what?

98: I believe this is the first place ‘GCP’ is used in the paper. This should be defined/introduced earlier.

179: unnecessary apostrophe in ‘points’

192: not clear what ‘high-water bed’ means

300: ‘know’ should be ‘known’

414: confused about the comparison of Agisoft and ODM RMSEs – with 9 GCPs, the bootstrap box plot appears to indicate that ODM has significantly larger error than Agisoft – approximately twice the error, with non-overlapping box plots. Yet the authors make a point that the results are comparable. Am I missing something? Having 2x error seems like a significant downside to me.

440-441: ‘The results indicate a decrease in the RMSE as we increase the number of GCPs’ – in general yes, but the RMSE increases when going from 5 to 9 GCPs before going down. Given 9 GCPs is presented as the optimal number to use, this seems problematic.

475: In Figure 14, there appears to be a substantial topographic artifact/mismatch in the upper part of the study area where the wet and dry bathymetry are merged, with noticeably higher elevations in the ‘wet’ bathymetry relative to the adjacent ‘dry’ bathymetry. This isn’t addressed in the paper, but seems like it may be an important issue, especially if this were to be used for hydraulic modeling at some point.

483-486: this text on FCP was already presented in the introduction.

504: it seems that the point of this slope analysis is to say that for these conditions the slope should not be estimated from the SfM data under any circumstances. Suggest saying something explicit like this when the data are presented.

512: ‘relatively’ – relatively what?

531: seems that the conveyance is more impacted by the quality of the ‘wet’ bathymetry collected by the GNSS than the ‘dry’ SfM bathymetry. This point could be emphasized a bit more when discussing the results.

Hide

RR by Salvatore Manfreda (02 Nov 2021)

ED: Publish subject to minor revisions (review by editor) (02 Nov 2021) by Antonio Annis

AR by Hubert Samboko on behalf of the Authors (12 Nov 2021) Author's response Author's tracked changes Manuscript

ED: Publish as is (22 Nov 2021) by Antonio Annis

AR by Hubert Samboko on behalf of the Authors (02 Dec 2021)

Short summary

The study was conducted along the Luangwa River in Zambia. It combines low-cost instruments such as UAVs and GPS kits to collect data for the purposes of water management. A novel technique which seamlessly merges the dry and wet bathymetry before application in a hydraulic model was applied. Successful implementation resulted in water authorities with small budgets being able to monitor flows safely and efficiently without significant compromise on accuracy.