the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Mar 2024
Effects of spatial resolution of digital terrain obtained by drone on mountainous urban fluvial flood modelling
Abstract. The effect of resolution and quality of terrain data, as the most sensitive input to 2D hydrodynamic modelling, has been one of the main research areas in flood modelling. However, previous studies have lacked the discussion on the limitation of the target area and the data source, as well as the underlying causes of simulation bias due to different resolutions. This study first discusses the performance of high-resolution DSM acquired by drone for flood modelling in a mountainous riverine city, and the effect of DSM resolution on results using grid resolutions from 6 cm to 30 m. The study then investigates the root causes of the effect based on topographic attributes. Xuanhan city, a riverine city in the mountainous region of southwest China, was used as the study area. The HEC-RAS 2D model was used for all simulations, and the results generated using 6 cm DSM acquired by drone were used as a benchmark. Results show that the simulation effect of flood characteristics shows a certain step change with the change of DSM resolution. DSMs with a resolution within 10 m can better capture the undulating features of the topography in the study area, which is crucial for the modelling of the inundation area. However, if features with specific elevation difference values are involved, it is best to keep the resolution within 5 m, which will have a direct impact on the accuracy of the modelling of the flood depth. The analysis of topographic attributes provides theoretical support for obtaining the optimal resolution to match simulation requirements.
- Preprint
(8717 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
CC1: 'Comment on egusphere-2024-404', Ziqi Yan, 28 Mar 2024
The application of high-resolution terrain data in flood forecasting and warning has been a hot research topic in this field. The most important difficulties in practical application are limitation of the cost of computing power and the operability.。The authors present a comprehensive exploration of drone-based acquisition of high-precision Digital Surface Models in the mountainous riverside city. The study delves into the influence of varying DSM resolutions on flood inundation simulations and investigates the fundamental factors driving these effects. The topic is interesting and the technical soundness of the method is reasonable, the contents are valuable for researchers and managers involved in the work. For better understanding of the methodology and application, some parts are still recommended to discussed and illustrated more, as follows:
- In Section 2.2, it is recommended to provide specific details regarding the setup of drone flight missions. This could include information on route planning and the precise positioning of control points. Additionally, in the results and discussion sections, it would be valuable to present the outcomes of drone aerial surveys (such as errors associated with ground control points) and the effectiveness of post-processing products. Such elaboration is crucial for readers to understand the workflow of drone operations and to establish the reliability of the result.
- When extracting and analyzing topographic attributes, the choice of using the average value within a grid cell as the central point’s indicator value is reasonable. The side length of grid cell was set at 30 m, and it is worth explaining the rationale behind this selection. If the goal is to ensure that each grid cell encompasses multiple complete raster pixels, opting for a larger side length than 30 m might be more reasonable
Citation: https://doi.org/10.5194/egusphere-2024-404-CC1 -
AC3: 'Reply on CC1', Xiaorong Huang, 23 Oct 2024
Author comments to community comment #1
1.“In Section 2.2, it is recommended to provide specific details regarding the setup of drone flight missions. This could include information on route planning and the precise positioning of control points. Additionally, in the results and discussion sections, it would be valuable to present the outcomes of drone aerial surveys (such as errors associated with ground control points) and the effectiveness of post-processing products. Such elaboration is crucial for readers to understand the workflow of drone operations and to establish the reliability of the result.”
Response: Thank you for your insightful comments.Since the figure already contains information about inundation points, we are concerned that adding control points may cause confusion. However, we will also include relevant content regarding control points.
2.“When extracting and analyzing topographic attributes, the choice of using the average value within a grid cell as the central point’s indicator value is reasonable. The side length of grid cell was set at 30 m, and it is worth explaining the rationale behind this selection. If the goal is to ensure that each grid cell encompasses multiple complete raster pixels, opting for a larger side length than 30 m might be more reasonable”
Response: Thank you for raising this important point. From the perspective of including multiple complete raster pixels, it may be more reasonable to choose a cell size larger than 30 meters. However, for flood simulations in mountainous riverside cities, a 30-meter resolution of topographic data is already too coarse. We still retain the discussion on 30-meter resolution as a point of comparison, as the currently mainstream DEMs (derived from satellite imagery) still use a 30-meter resolution. Our main focus of discussion remains on the range of resolutions between 6 cm and 15 m.
Thank you for taking the time to review our paper and providing such valuable feedback.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC3
-
RC1: 'Comment on egusphere-2024-404', Anonymous Referee #1, 18 Apr 2024
Summary: This study presents a 2D flood inundation modeling study in China. Authors investigate how the quality of flood inundation simulations depends on the resolution of the underlying digital elevation model. A high-resolution (6 cm ground resolution) DEM is derived from airborne imagery and flood simulations are performed at different resolutions ranging from 6 cm to 30 m. Results presented in this paper are interesting for the flood inundation modeling community and I recommend publication of this article after revisions, as outlined below.
Review comments:
- Improve language and grammar. The text is not always easy to read. Use ChatGPT, Google gemini or similar to improve your text, it will make the paper more accessible and impactful.
- Authors have used photogrammetry techniques to derive the DEM from drone-borne imagery. Another option would be drone-borne lidar. I think the paper would benefit from a short summary of these available options, pros and cons of each option and a few references illustrating those.
- A key step for hydraulic modeling purposes is generating the terrain model from the surface model, i.e. DSM to DTM conversion. Main factors in this are buildings and vegetation. The article explains that buildings were not removed, which I agree with – how was vegetation handled?
- As pointed out in the paper, submerged topography cannot be derived from photogrammetric DEMs. However, there are options to get bathymetric information from UAS, including green lidar, water penetrating radar and sonar. Those recent advances could be briefly summarized in the paper.
- A key result of the paper is to show how the quality of flood simulations depends on the DEM resolution. It is perhaps not surprising that quality increases with increasing resolution. One key factor that limits what is possible, at least for larger domains, is CPU time and computing resources. It would thus be informative to see those parameters for the different model runs at different spatial resolution.
- Authors introduce and calculate quantitative terrain indicators called topographic features here. They show that these indicators depend on the spatial resolution of the DEM and that the mismatch between the features calculated at native and coarsened resolution increases with decreasing resolution. Further, they argue that degradation of the terrain indicators with decreasing resolution is similar to the degradation of the flood simulation results with decreasing resolution. It would be good to analyze this further: What is the correlation between different topographic features and skill of the flood simulation? Are there specific thresholds for the feature mismatches, exceedance of which would cause the skill of the flood simulation to decrease? Are these findings site-independent and transferable?
Details:
- Avoid acronyms in abstract or write out before first use (DSM)
- Line 17: Replace “within” with “better than”
- Line 18-20: Please rephrase sentence “However…. flood depth”. Unclear
- Line 29: El Nino is not induced by global warming – rephrase
- Line 50: Be consistent with the terms DSM, DEM and DTM. DTM is the main input to flood models.
- Line 141 ff: There are drone-borne bathymetry options (https://doi.org/10.5194/hess-22-4165-2018, https://doi.org/10.1016/j.jhydrol.2022.128789)
- Line 204: Which model parameters?
- Fig 7: Legend items are mis-spelled
- Fig 9: Provide units for y-axis (absolute error)
Citation: https://doi.org/10.5194/egusphere-2024-404-RC1 -
AC1: 'Reply on RC1', Xiaorong Huang, 23 Oct 2024
Author Comments to Referee #1
Responses to the general review comments:
1.“Improve language and grammar. The text is not always easy to read. Use ChatGPT, Google gemini or similar to improve your text, it will make the paper more accessible and impactful.”
Response: Thanks very much for taking your time to review our manuscript. We have carefully checked and improved the English writing in the revised manuscript, and we believe the readers could understand our work more clearly.
2. “Authors have used photogrammetry techniques to derive the DEM from drone-borne imagery. Another option would be drone-borne lidar. I think the paper would benefit from a short summary of these available options, pros and cons of each option and a few references illustrating those.”
Response: Thank you for your kind suggestion. We fully agree to include a brief summary of other DEM acquisition methods based on drones, and this content will be added to the introduction.
3.“A key step for hydraulic modeling purposes is generating the terrain model from the surface model, i.e. DSM to DTM conversion. Main factors in this are buildings and vegetation. The article explains that buildings were not removed, which I agree with – how was vegetation handled?”
Response: Thank you for your insightful comments. We processed the DSM using the PCI Geomatica software. The reason for choosing PCI Geomatica is that it allows for manual local editing of the DSM, which is a better choice for our study area as the vegetation zones are relatively small and scattered. Compared to applying global filtering to the entire study area, local processing is more suitable. The specific processing method involves using various filters available in PCI Geomatica, such as Terrain filters, Pit and Bump Filters, Median filters, and Clamp filters, along with manual touch-up edits to filter the designated vegetation areas. PCI Geomatica can also take into account the slope of the local terrain and apply directional filtering to more accurately remove vegetation and other non-ground features.We will supplement the method section with the specific formulas used by the filters and clarify which filters were applied in this study.
4.“As pointed out in the paper, submerged topography cannot be derived from photogrammetric DEMs. However, there are options to get bathymetric information from UAS, including green lidar, water penetrating radar and sonar. Those recent advances could be briefly summarized in the paper.”
Response: Thanks for your professional suggestions. During our research, we also discussed the current progress in obtaining water depth information and underwater topography using drones with experts in drone surveying. Some conclusions were given at that time, such as the fact that the performance of green LiDAR is highly dependent on water clarity, and integrating green LiDAR data with other datasets (e.g., traditional LiDAR or sonar) may be complex.We will add this missing research progress explanation to the introduction.
5.“A key result of the paper is to show how the quality of flood simulations depends on the DEM resolution. It is perhaps not surprising that quality increases with increasing resolution. One key factor that limits what is possible, at least for larger domains, is CPU time and computing resources. It would thus be informative to see those parameters for the different model runs at different spatial resolution.”
Response: Thank you for raising this important point. CPU time and computing resources are important information in this study, as improving computational efficiency is one of the research significances of this paper. We will supplement the hardware information of all the computers used for simulations and the computation time for processing data at different resolutions, which will allow readers to more intuitively perceive the changes in computational efficiency.
6.“Authors introduce and calculate quantitative terrain indicators called topographic features here. They show that these indicators depend on the spatial resolution of the DEM and that the mismatch between the features calculated at native and coarsened resolution increases with decreasing resolution. Further, they argue that degradation of the terrain indicators with decreasing resolution is similar to the degradation of the flood simulation results with decreasing resolution. It would be good to analyze this further: What is the correlation between different topographic features and skill of the flood simulation? Are there specific thresholds for the feature mismatches, exceedance of which would cause the skill of the flood simulation to decrease? Are these findings site-independent and transferable?”
Response: We would like to thank the reviewer for these constructive comments.This study uses the statement "the accuracy of terrain undulation representation varies with different resolutions, thereby affecting the accuracy of flood simulation" to qualitatively describe the correlation between various terrain features and flood simulation techniques. How these factors are specifically related is a question we have been seeking to resolve. We believe this connection is related to the equations used for calculating two-dimensional unsteady flow in hydraulic models. Various topographic features are directly linked to the terrain gradient terms in the hydraulic equations, influencing changes in water momentum, flow velocity, and direction. This requires further in-depth research and discussion in our future work.
Regarding whether there is a specific threshold for feature mismatch, this is also a question we aim to address in this study. Our original research plan was to design a scoring system based on topographic attribute metrics, improving the statistical topographic attribute metrics into a comprehensive evaluation index. We would then use finer resolution intervals to comprehensively score the accuracy of terrain undulation representation under different resolutions, while also incorporating flood inundation simulation accuracy and computational cost into the evaluation system. This would not only facilitate discussion on whether a specific threshold exists but also help discover how these methods apply in different regions. We are currently making efforts to find suitable improvement methods.
Responses to the minor comments:
1.Avoid acronyms in abstract or write out before first use (DSM)
Response: Thank you for pointing out the problem, we will modify the abbreviation.
2.Line 17: Replace “within” with “better than”
Response:Thanks for your comment.We will revise the expression.
3.Line 18-20: Please rephrase sentence “However…. flood depth”. Unclear
Response:Thanks for your comment.We will revise the expression.
4.Line 29: El Nino is not induced by global warming – rephrase
Response:Thank you for pointing out this important expression error, we will correct the expression.
5.Line 50: Be consistent with the terms DSM, DEM and DTM. DTM is the main input to flood models.
Response:We fully agree with the reviewer’s opinion, and we will revise the corresponding expressions accordingly.
6.Line 141: There are drone-borne bathymetry options (https://doi.org/10.5194/hess-22-4165-2018, https://doi.org/10.1016/j.jhydrol.2022.128789)
Response: We greatly appreciate the reviewer for providing the references. We will carefully review them and include additional explanations regarding UAV-based bathymetry in the introduction.
7.Line 204: Which model parameters?
Response: The parameters adjusted in the hydraulic model are the Manning's roughness coefficient and the channel slope.
8.Fig 7: Legend items are mis-spelled
Response: We sincerely appreciate the reviewer for pointing out the spelling errors. We will correct the expressions in the figure legend accordingly.
9.Fig 9: Provide units for y-axis (absolute error)
Response: We are grateful to the reviewer for pointing out this issue. We will revise the result figures for the indicators that include units accordingly.
We sincerely appreciate the reviewer’s comments on the details of the manuscript, especially for pointing out some errors and confusions in the expressions. We will carefully revise the relevant content in the next version of the manuscript. Once again, thank you for taking the time to review our paper and providing such valuable feedback.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC1
-
RC2: 'Comment on egusphere-2024-404', Anonymous Referee #2, 20 Oct 2024
The authors present an interesting paper in which they carry out a high resolution hydrodynamic modeling of a mountainous riverine city and aim to investigate and discuss the effect of increasing DSM resolution (from 6 cm to 30 m) on the hydrodynamic model results. The paper is well written and can be considered fr publication only after some major revisions, which are suggested and detailed below.
Major comments
Being the topography one crucial element of the paper there is a need to define once in the text the definition of dsm and dtm, their differences and which one of them is used in the model. The post processing of the dsm should be detailed in order to make it clear and reproducible. The authors should move from spelling the name of the software used to spelling the single steps of the procedure (and the main algorithms) they used. Is there any computation of levees lines, special topographic features/lines, artificial depressions filling?
The authors often acknowledge the fact that the study area is mountainous and urbanized. However there is not reference, no discussion and no comparison against studies that include physical processes such as sediment transport and turbidity in the flood modeling as well as concerning the urban drainage network. What the assumptions used, and the hypothesis made by the authors concerning these (neglected) concepts? I believe this should be clarified and offer interesting points to be discusses.
The validation procedure is not clear. Two concepts should be clarified: i) inundation points: the authors should spell out if they are points, cross sections, groups of flooded pixels. Are they flooded by a single or multiple flood events? Are those measured localized data? ii) the authors use the 6cm simulation as benchmark, but no spatial error metrics are presented over the analyzed domain. This is crucial for having an overall model assessment with varying resolution and not only at the 6 points.
Minor comments:
- Figure 1: Inundation points and the corresponding text are not readable from the insert of Figure 1 please use a different color
- Table 1: serial number is not a number but a letter please be consistent in the table description and in the text.
- Line 88: simulation errors: errors or dissimilarities, if errors please clarify how these errors are computed, against what type of ground truth?
- Line 116: please clarify what is the meaning of “typycal” and what are these main feautures
- Line 116: According to the 2022 Flood Control Plan of Xuanhan: I wonder if the domain was already used in the According to the 2022 Flood Control Plan of Xuanhan and if inundation maps are available in the plan or by local authorities over the same study area. This will provide an independent validation of the method used in this paper, which move from 6 points to spatial assessment.
- Line 118: what is a warning points, how is this defined? in a location/a cross section, an area of pixels well defined where inundation was observed? if yes, what about the surrounding of the 6 points?
- Line 121: please be more explicit on how did you select the flood event (if it was only one or more than one); is this measured or simulated? if simulated by which kind of model? how about initial hydrological/hydraulic conditions that generated this event?
- Line 134: can you better define the PCI Geomatica, which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
- Line 137: carried out using the highest quality settings: can you please specify what does it means? which processes are involved in the “high quality“ and how “high“ is defined?
- Line 148: processing tool RAS Mapper: which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
- Line 150: a table could be useful where the authors clearly mention all the input, the parameters for setting the model simulations, and all the output. This should clarify the procedure and eventually the study reproducibility, which is crucial in research.
- Lines 156-162 please clarify the meaning, the units, and values of the variables presented
- Line 162 radius: please space is missing
- Line 173: what is a flood process: do you mean a flood hydrograph?
- Line 181: how and why did you use these attributes? the meaning of each of them is clear but how and why they will be used in the methodology is not clear.
- Line 201-204: this is not results, it is method; still not clear what are the 6 points
- Line 208 210: what physically happen in correspondence of these six points/cross sections? change of slope, drainage directions? roughness?
- Figure 7: what are the dashed lines in panel b? please define them
- Line 224-225 please the sentence does not make sense, consider to revise or remove the .
- Line 225: "The reason for the lack of significant trends " do you mean statistically significant or only a clear pattern? please clarify
- Line 258: remove 1 dot
- Line 266, please define what the numbers in table 3 mean: eventually with a formula. In the caption mean error is used but how is it computed and why average error is considered; can you add more typical measures such as rmse, absolute error, percentage bias?
Citation: https://doi.org/10.5194/egusphere-2024-404-RC2 -
AC2: 'Reply on RC2', Xiaorong Huang, 23 Oct 2024
Author Comments to Referee#2
Responses to the general review comments:
1.“Being the topography one crucial element of the paper there is a need to define once in the text the definition of dsm and dtm, their differences and which one of them is used in the model. The post processing of the dsm should be detailed in order to make it clear and reproducible. The authors should move from spelling the name of the software used to spelling the single steps of the procedure (and the main algorithms) they used. Is there any computation of levees lines, special topographic features/lines, artificial depressions filling?”
Response: Thank you for your insightful comments. Sorry for the confusion in our expression regarding DSM and DTM. In most past studies, DTM (which focuses only on the natural terrain without surface features like trees or buildings) was used as the input for hydraulic models. However, in this study, since it involves flood inundation simulation, we retained the riverside buildings while processing the DSM obtained from the drone (DSM includes all surface objects), and only filtered out noise points from vegetation, water surfaces, and roads. As a result, the terrain data is in a state between DSM and DTM. After discussion and considering other reviewers' suggestions, we have decided that it is more appropriate to define the terrain data we used as DTM. We will supplement the relevant modifications and explanations in the section on processing drone imagery.
The specific processing method of DSM involves using various filters available in PCI Geomatica software, such as Terrain filters, Pit and Bump Filters, Median filters, and Clamp filters, along with manual touch-up edits to filter the designated vegetation areas.The specific process of handling DSM in this study is indeed important for the readers. Previously, we did not provide detailed explanations due to space constraints. We will supplement the method section with the specific formulas used by the filters and clarify which filters were applied in this study.
When using HEC-RAS for hydraulic simulation, we did not set additional levee lines or other special topographic features/lines, because, unlike traditional low-resolution DEMs, the 6 cm resolution DSM can accurately and realistically reflect terrain obstacles (such as levees and highlands). This was also verified during the cross-section processing and subsequent simulations in HEC-RAS. As for the artificial depressions filling, we handled it using Pit and Bump Filters during the DSM processing.
2. “The authors often acknowledge the fact that the study area is mountainous and urbanized. However there is not reference, no discussion and no comparison against studies that include physical processes such as sediment transport and turbidity in the flood modeling as well as concerning the urban drainage network. What the assumptions used, and the hypothesis made by the authors concerning these (neglected) concepts? I believe this should be clarified and offer interesting points to be discusses.”
Response: Thanks for your professional suggestions. We apologize for not providing a more detailed explanation regarding the impact of sediment transport and urban drainage networks in the manuscript. The reason we did not consider the influence of sediment transport in this study is due to the fact that the research area is located immediately downstream of a large reservoir, which intercepts sediment and results in the release of clear water. Additionally, we referred to the situation in many flood-prone mountainous riverside cities, where—except for extremely underdeveloped areas—reservoirs or other water-retaining structures are typically built upstream to mitigate the effects of flood disasters. These structures also reduce the downstream impact of sediment transport, even during flood discharges, as they provide a degree of sediment interception. Furthermore, since our research focuses on the main urban river section most affected by floods, without considering the broader flood propagation, we concluded that sediment transport has a limited impact on the flood simulation in our study.
As for the impact of urban drainage networks, as mentioned in the introduction of the manuscript, unlike plain cities where urban flooding is mainly caused by impervious surfaces and drainage networks, riverside cities in mountainous areas have impervious roads and various buildings constructed along significantly sloped terrains. During heavy rainfall, runoff rapidly converges towards the lowest areas of the river channel, making it difficult for significant urban flooding to occur. Additionally, flash floods in such areas are dominated by fluvial flood, with the inundation range gradually spreading along both sides of the river. The main human and economic damages occur close to the riverbanks. The rapid rise and fall characteristics of flash floods mean that the processes of flood initiation, peak passage, and recession could all be completed before the urban drainage network in mountainous cities has any significant impact. The main factors influencing flood propagation in such areas are the downstream channel’s flow capacity and upstream inflow variations, with the influence of the urban drainage network being relatively minor in comparison to these factors.
We will add the corresponding explanation to the Materials and Methods section, and we would like to once again thank you for pointing out these key issues.
3.“The validation procedure is not clear. Two concepts should be clarified: i) inundation points: the authors should spell out if they are points, cross sections, groups of flooded pixels. Are they flooded by a single or multiple flood events? Are those measured localized data? ii) the authors use the 6cm simulation as benchmark, but no spatial error metrics are presented over the analyzed domain. This is crucial for having an overall model assessment with varying resolution and not only at the 6 points.”
Response: Thank you for raising these important points. We sincerely apologize for the lack of clarity in the validation procedure. In fact, the term "inundation point" refers to a specific location indicating the flood boundary, which in reality is a flood boundary line. This line represents the position that the flood reaches when the water flow attains a certain value. These six flood boundary locations were selected by the local flood management department as key flood warning and observation points. During the flood season, staff members are stationed at these points to observe conditions and oversee evacuation procedures.
The flood boundary lines for each location were derived from historical observations over multiple flood events, which are summarized in the form of red lines (as shown in the field investigation photos in Figure 5). This data was provided by the technical staff of the Dazhou Hydrological and Water Resources Survey Centre (one of the co-authors of this study). During our field investigation, we also verified these lines through the inspection of historical flood traces and interviews with local residents.We will include this additional explanation in the relevant sections of the manuscript.
Regarding the issue of missing spatial error metrics, we fully agree with the reviewer’s opinion. For the entire study area, we only used inundation area and average inundation depth as evaluation metrics, which are indeed insufficient for a comprehensive assessment of the model. Moving forward, we plan to include additional metrics to statistically evaluate the differences in flood depth and inundation extent simulations.
Responses to the minor comments:
1.Figure 1: Inundation points and the corresponding text are not readable from the insert of Figure 1 please use a different color
Response:We appreciate the reviewer for pointing out this issue. We will revise the colors of the inundation points in Figure 1 accordingly.
2.Table 1: serial number is not a number but a letter please be consistent in the table description and in the text.
Response: Thank for your suggestion. We will correct the erroneous statements in Table 1.
3.Line 88: simulation errors: errors or dissimilarities, if errors please clarify how these errors are computed, against what type of ground truth?
Response: Thanks for your comment. We will supplement the calculation methods for the errors and the types of ground truth used for comparison, along with the corresponding references.
4.Line 116: please clarify what is the meaning of “typycal” and what are these main feautures
Response:Thanks for your comment. We will provide additional explanations for the features of typical mountainous riverside cities, including Geographical Location, Topography, and Urban Distribution.
5.Line 116: According to the 2022 Flood Control Plan of Xuanhan: I wonder if the domain was already used in the According to the 2022 Flood Control Plan of Xuanhan and if inundation maps are available in the plan or by local authorities over the same study area. This will provide an independent validation of the method used in this paper, which move from 6 points to spatial assessment.
Response:Thanks for your comment.Relevant explanations can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
6.Line 118: what is a warning points, how is this defined? in a location/a cross section, an area of pixels well defined where inundation was observed? if yes, what about the surrounding of the 6 points?
Response:Thanks for your comment.Relevant explanations can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
7.Line 121: please be more explicit on how did you select the flood event (if it was only one or more than one); is this measured or simulated? if simulated by which kind of model? how about initial hydrological/hydraulic conditions that generated this event?
Response:Thanks for your comment. We used observational data from the major flood event in 2005, and we will revise the unclear expression.
8.Line 134: can you better define the PCI Geomatica, which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
Response:Thanks for your comment.Relevant explanations can be found in the response to the first major comment above, and we will also include additional clarifications in the manuscript.
9.Line 137: carried out using the highest quality settings: can you please specify what does it means? which processes are involved in the “high quality“ and how “high“ is defined?
Response:Thanks for your comment.The term "high quality" refers to an option in DJI's drone imaging processing software, DJI Terra. High-quality processing signifies the generation of denser and more refined point cloud data, utilizing advanced noise filtering techniques, among others. This approach places higher demands on both image quality and computational hardware. We will provide corresponding explanations in the manuscript.
10.Line 148: processing tool RAS Mapper: which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
Response:Thanks for your comment.One of the major problems in hydraulic modeling is that terrain data does not often include the actual terrain underneath the water surface in the channel region. RAS Mapper can now be used to create a terrain model of the channel region from the HECRAS cross sections and the Cross Section Interpolation Surface. This terrain model can then be combined with the general surface terrain model (that does not accurately depict the terrain below the water surface) to create an improved terrain model for hydraulic modeling and mapping.
11.Line 150: a table could be useful where the authors clearly mention all the input, the parameters for setting the model simulations, and all the output. This should clarify the procedure and eventually the study reproducibility, which is crucial in research.
Response:Thanks for your comment.We will include a table in the manuscript that contains all the inputs, the parameters used for setting up the model simulations, and all the outputs.
12.Lines 156-162 please clarify the meaning, the units, and values of the variables presented
Response:Thanks for your comment.We will supplement the meaning, units, and values of the variables presented.
13.Line 162 radius: please space is missing
Response:Thanks for your comment.We will correct the missing spaces.
14.Line 173: what is a flood process: do you mean a flood hydrograph?
Response:Thanks for your comment.This refers to using the runoff time series of the selected flood events as input. We will revise the wording that caused the confusion.
15.Line 181: how and why did you use these attributes? the meaning of each of them is clear
but how and why they will be used in the methodology is not clear.
Response:Thanks for your comment.These six attributes comprehensively account for the factors of elevation, terrain ruggedness, exposure, and morphometric protection, as well as flow resistance, providing a thorough evaluation of how topographic features influence hydrodynamics. For instance:
Elevation: Elevation is a critical topographic parameter affecting water flow speed, direction, and energy. It directly determines the downhill flow tendency, making it essential for flood simulation.
TPI (Topographic Position Index): TPI describes the position of a point within its surrounding terrain (e.g., hilltops or valleys). It helps determine areas where water collects or disperses, influencing flow paths and ponding locations.
TRI (Terrain Ruggedness Index): TRI measures terrain roughness, describing the degree of surface undulation. Rugged terrain can increase flow friction, affecting both flow velocity and water kinetic energy.
MPI (Morphometric Protection Index): MPI reflects the protective role of terrain against natural disasters like landslides or erosion. In hydraulic models, this index helps assess the influence of terrain on flow paths.
VRM (Vector Ruggedness Measure): VRM is another measure of terrain complexity, capturing surface variation in all directions. Complex terrain can significantly alter water flow paths and velocities, and in flood simulations, VRM assists in identifying areas where water flow might be obstructed.
We will incorporate this clarification into the Table 2.
16.Line 201-204: this is not results, it is method; still not clear what are the 6 points
Response:Thanks for your comment. We will supplement and expand the content of this section, and the explanation regarding the inundation points can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
17.Line 208 210: what physically happen in correspondence of these six points/cross sections? change of slope, drainage directions? roughness?
Response:Thanks for your comment. The explanation regarding the inundation points can be found in the response to the third major comment above.
18.Figure 7: what are the dashed lines in panel b? please define them
Response:Thanks for your comment.The dashed lines are used to better illustrate the variations (step changes) between the simulation results at different resolutions. We will add relevant explanations in the manuscript.
19.Line 224-225 please the sentence does not make sense, consider to revise or remove the .
Response:Thanks for your comment.We will refine the expression.
20.Line 225: "The reason for the lack of significant trends " do you mean statistically significant or only a clear pattern? please clarify
Response:Thanks for your comment.The term "not significant" here refers to the observation of a clear pattern derived solely from the bar charts in the figure, as it only involves results from six different resolutions (which provides a limited statistical sample), and there are no apparent differences in the results.
21.Line 258: remove 1 dot
Response:Thanks for your comment.We will correct this error.
22 .Line 266, please define what the numbers in table 3 mean: eventually with a formula. In the caption mean error is used but how is it computed and why average error is considered; can you add more typical measures such as RMSE, absolute error, percentage bias?
Response:Thanks for your professional suggestions. We will revise the results in Table 3 and their corresponding descriptions, and we will also add other typical measures.
Thank you very much for taking the time to review our manuscript and for providing many professional evaluations. Each comment has been extremely helpful in enhancing the professionalism and readability of the manuscript. We will carefully revise the manuscript according to your suggestions. Once again, we appreciate the valuable feedback provided by the reviewer.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC2
Status: closed
-
CC1: 'Comment on egusphere-2024-404', Ziqi Yan, 28 Mar 2024
The application of high-resolution terrain data in flood forecasting and warning has been a hot research topic in this field. The most important difficulties in practical application are limitation of the cost of computing power and the operability.。The authors present a comprehensive exploration of drone-based acquisition of high-precision Digital Surface Models in the mountainous riverside city. The study delves into the influence of varying DSM resolutions on flood inundation simulations and investigates the fundamental factors driving these effects. The topic is interesting and the technical soundness of the method is reasonable, the contents are valuable for researchers and managers involved in the work. For better understanding of the methodology and application, some parts are still recommended to discussed and illustrated more, as follows:
- In Section 2.2, it is recommended to provide specific details regarding the setup of drone flight missions. This could include information on route planning and the precise positioning of control points. Additionally, in the results and discussion sections, it would be valuable to present the outcomes of drone aerial surveys (such as errors associated with ground control points) and the effectiveness of post-processing products. Such elaboration is crucial for readers to understand the workflow of drone operations and to establish the reliability of the result.
- When extracting and analyzing topographic attributes, the choice of using the average value within a grid cell as the central point’s indicator value is reasonable. The side length of grid cell was set at 30 m, and it is worth explaining the rationale behind this selection. If the goal is to ensure that each grid cell encompasses multiple complete raster pixels, opting for a larger side length than 30 m might be more reasonable
Citation: https://doi.org/10.5194/egusphere-2024-404-CC1 -
AC3: 'Reply on CC1', Xiaorong Huang, 23 Oct 2024
Author comments to community comment #1
1.“In Section 2.2, it is recommended to provide specific details regarding the setup of drone flight missions. This could include information on route planning and the precise positioning of control points. Additionally, in the results and discussion sections, it would be valuable to present the outcomes of drone aerial surveys (such as errors associated with ground control points) and the effectiveness of post-processing products. Such elaboration is crucial for readers to understand the workflow of drone operations and to establish the reliability of the result.”
Response: Thank you for your insightful comments.Since the figure already contains information about inundation points, we are concerned that adding control points may cause confusion. However, we will also include relevant content regarding control points.
2.“When extracting and analyzing topographic attributes, the choice of using the average value within a grid cell as the central point’s indicator value is reasonable. The side length of grid cell was set at 30 m, and it is worth explaining the rationale behind this selection. If the goal is to ensure that each grid cell encompasses multiple complete raster pixels, opting for a larger side length than 30 m might be more reasonable”
Response: Thank you for raising this important point. From the perspective of including multiple complete raster pixels, it may be more reasonable to choose a cell size larger than 30 meters. However, for flood simulations in mountainous riverside cities, a 30-meter resolution of topographic data is already too coarse. We still retain the discussion on 30-meter resolution as a point of comparison, as the currently mainstream DEMs (derived from satellite imagery) still use a 30-meter resolution. Our main focus of discussion remains on the range of resolutions between 6 cm and 15 m.
Thank you for taking the time to review our paper and providing such valuable feedback.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC3
-
RC1: 'Comment on egusphere-2024-404', Anonymous Referee #1, 18 Apr 2024
Summary: This study presents a 2D flood inundation modeling study in China. Authors investigate how the quality of flood inundation simulations depends on the resolution of the underlying digital elevation model. A high-resolution (6 cm ground resolution) DEM is derived from airborne imagery and flood simulations are performed at different resolutions ranging from 6 cm to 30 m. Results presented in this paper are interesting for the flood inundation modeling community and I recommend publication of this article after revisions, as outlined below.
Review comments:
- Improve language and grammar. The text is not always easy to read. Use ChatGPT, Google gemini or similar to improve your text, it will make the paper more accessible and impactful.
- Authors have used photogrammetry techniques to derive the DEM from drone-borne imagery. Another option would be drone-borne lidar. I think the paper would benefit from a short summary of these available options, pros and cons of each option and a few references illustrating those.
- A key step for hydraulic modeling purposes is generating the terrain model from the surface model, i.e. DSM to DTM conversion. Main factors in this are buildings and vegetation. The article explains that buildings were not removed, which I agree with – how was vegetation handled?
- As pointed out in the paper, submerged topography cannot be derived from photogrammetric DEMs. However, there are options to get bathymetric information from UAS, including green lidar, water penetrating radar and sonar. Those recent advances could be briefly summarized in the paper.
- A key result of the paper is to show how the quality of flood simulations depends on the DEM resolution. It is perhaps not surprising that quality increases with increasing resolution. One key factor that limits what is possible, at least for larger domains, is CPU time and computing resources. It would thus be informative to see those parameters for the different model runs at different spatial resolution.
- Authors introduce and calculate quantitative terrain indicators called topographic features here. They show that these indicators depend on the spatial resolution of the DEM and that the mismatch between the features calculated at native and coarsened resolution increases with decreasing resolution. Further, they argue that degradation of the terrain indicators with decreasing resolution is similar to the degradation of the flood simulation results with decreasing resolution. It would be good to analyze this further: What is the correlation between different topographic features and skill of the flood simulation? Are there specific thresholds for the feature mismatches, exceedance of which would cause the skill of the flood simulation to decrease? Are these findings site-independent and transferable?
Details:
- Avoid acronyms in abstract or write out before first use (DSM)
- Line 17: Replace “within” with “better than”
- Line 18-20: Please rephrase sentence “However…. flood depth”. Unclear
- Line 29: El Nino is not induced by global warming – rephrase
- Line 50: Be consistent with the terms DSM, DEM and DTM. DTM is the main input to flood models.
- Line 141 ff: There are drone-borne bathymetry options (https://doi.org/10.5194/hess-22-4165-2018, https://doi.org/10.1016/j.jhydrol.2022.128789)
- Line 204: Which model parameters?
- Fig 7: Legend items are mis-spelled
- Fig 9: Provide units for y-axis (absolute error)
Citation: https://doi.org/10.5194/egusphere-2024-404-RC1 -
AC1: 'Reply on RC1', Xiaorong Huang, 23 Oct 2024
Author Comments to Referee #1
Responses to the general review comments:
1.“Improve language and grammar. The text is not always easy to read. Use ChatGPT, Google gemini or similar to improve your text, it will make the paper more accessible and impactful.”
Response: Thanks very much for taking your time to review our manuscript. We have carefully checked and improved the English writing in the revised manuscript, and we believe the readers could understand our work more clearly.
2. “Authors have used photogrammetry techniques to derive the DEM from drone-borne imagery. Another option would be drone-borne lidar. I think the paper would benefit from a short summary of these available options, pros and cons of each option and a few references illustrating those.”
Response: Thank you for your kind suggestion. We fully agree to include a brief summary of other DEM acquisition methods based on drones, and this content will be added to the introduction.
3.“A key step for hydraulic modeling purposes is generating the terrain model from the surface model, i.e. DSM to DTM conversion. Main factors in this are buildings and vegetation. The article explains that buildings were not removed, which I agree with – how was vegetation handled?”
Response: Thank you for your insightful comments. We processed the DSM using the PCI Geomatica software. The reason for choosing PCI Geomatica is that it allows for manual local editing of the DSM, which is a better choice for our study area as the vegetation zones are relatively small and scattered. Compared to applying global filtering to the entire study area, local processing is more suitable. The specific processing method involves using various filters available in PCI Geomatica, such as Terrain filters, Pit and Bump Filters, Median filters, and Clamp filters, along with manual touch-up edits to filter the designated vegetation areas. PCI Geomatica can also take into account the slope of the local terrain and apply directional filtering to more accurately remove vegetation and other non-ground features.We will supplement the method section with the specific formulas used by the filters and clarify which filters were applied in this study.
4.“As pointed out in the paper, submerged topography cannot be derived from photogrammetric DEMs. However, there are options to get bathymetric information from UAS, including green lidar, water penetrating radar and sonar. Those recent advances could be briefly summarized in the paper.”
Response: Thanks for your professional suggestions. During our research, we also discussed the current progress in obtaining water depth information and underwater topography using drones with experts in drone surveying. Some conclusions were given at that time, such as the fact that the performance of green LiDAR is highly dependent on water clarity, and integrating green LiDAR data with other datasets (e.g., traditional LiDAR or sonar) may be complex.We will add this missing research progress explanation to the introduction.
5.“A key result of the paper is to show how the quality of flood simulations depends on the DEM resolution. It is perhaps not surprising that quality increases with increasing resolution. One key factor that limits what is possible, at least for larger domains, is CPU time and computing resources. It would thus be informative to see those parameters for the different model runs at different spatial resolution.”
Response: Thank you for raising this important point. CPU time and computing resources are important information in this study, as improving computational efficiency is one of the research significances of this paper. We will supplement the hardware information of all the computers used for simulations and the computation time for processing data at different resolutions, which will allow readers to more intuitively perceive the changes in computational efficiency.
6.“Authors introduce and calculate quantitative terrain indicators called topographic features here. They show that these indicators depend on the spatial resolution of the DEM and that the mismatch between the features calculated at native and coarsened resolution increases with decreasing resolution. Further, they argue that degradation of the terrain indicators with decreasing resolution is similar to the degradation of the flood simulation results with decreasing resolution. It would be good to analyze this further: What is the correlation between different topographic features and skill of the flood simulation? Are there specific thresholds for the feature mismatches, exceedance of which would cause the skill of the flood simulation to decrease? Are these findings site-independent and transferable?”
Response: We would like to thank the reviewer for these constructive comments.This study uses the statement "the accuracy of terrain undulation representation varies with different resolutions, thereby affecting the accuracy of flood simulation" to qualitatively describe the correlation between various terrain features and flood simulation techniques. How these factors are specifically related is a question we have been seeking to resolve. We believe this connection is related to the equations used for calculating two-dimensional unsteady flow in hydraulic models. Various topographic features are directly linked to the terrain gradient terms in the hydraulic equations, influencing changes in water momentum, flow velocity, and direction. This requires further in-depth research and discussion in our future work.
Regarding whether there is a specific threshold for feature mismatch, this is also a question we aim to address in this study. Our original research plan was to design a scoring system based on topographic attribute metrics, improving the statistical topographic attribute metrics into a comprehensive evaluation index. We would then use finer resolution intervals to comprehensively score the accuracy of terrain undulation representation under different resolutions, while also incorporating flood inundation simulation accuracy and computational cost into the evaluation system. This would not only facilitate discussion on whether a specific threshold exists but also help discover how these methods apply in different regions. We are currently making efforts to find suitable improvement methods.
Responses to the minor comments:
1.Avoid acronyms in abstract or write out before first use (DSM)
Response: Thank you for pointing out the problem, we will modify the abbreviation.
2.Line 17: Replace “within” with “better than”
Response:Thanks for your comment.We will revise the expression.
3.Line 18-20: Please rephrase sentence “However…. flood depth”. Unclear
Response:Thanks for your comment.We will revise the expression.
4.Line 29: El Nino is not induced by global warming – rephrase
Response:Thank you for pointing out this important expression error, we will correct the expression.
5.Line 50: Be consistent with the terms DSM, DEM and DTM. DTM is the main input to flood models.
Response:We fully agree with the reviewer’s opinion, and we will revise the corresponding expressions accordingly.
6.Line 141: There are drone-borne bathymetry options (https://doi.org/10.5194/hess-22-4165-2018, https://doi.org/10.1016/j.jhydrol.2022.128789)
Response: We greatly appreciate the reviewer for providing the references. We will carefully review them and include additional explanations regarding UAV-based bathymetry in the introduction.
7.Line 204: Which model parameters?
Response: The parameters adjusted in the hydraulic model are the Manning's roughness coefficient and the channel slope.
8.Fig 7: Legend items are mis-spelled
Response: We sincerely appreciate the reviewer for pointing out the spelling errors. We will correct the expressions in the figure legend accordingly.
9.Fig 9: Provide units for y-axis (absolute error)
Response: We are grateful to the reviewer for pointing out this issue. We will revise the result figures for the indicators that include units accordingly.
We sincerely appreciate the reviewer’s comments on the details of the manuscript, especially for pointing out some errors and confusions in the expressions. We will carefully revise the relevant content in the next version of the manuscript. Once again, thank you for taking the time to review our paper and providing such valuable feedback.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC1
-
RC2: 'Comment on egusphere-2024-404', Anonymous Referee #2, 20 Oct 2024
The authors present an interesting paper in which they carry out a high resolution hydrodynamic modeling of a mountainous riverine city and aim to investigate and discuss the effect of increasing DSM resolution (from 6 cm to 30 m) on the hydrodynamic model results. The paper is well written and can be considered fr publication only after some major revisions, which are suggested and detailed below.
Major comments
Being the topography one crucial element of the paper there is a need to define once in the text the definition of dsm and dtm, their differences and which one of them is used in the model. The post processing of the dsm should be detailed in order to make it clear and reproducible. The authors should move from spelling the name of the software used to spelling the single steps of the procedure (and the main algorithms) they used. Is there any computation of levees lines, special topographic features/lines, artificial depressions filling?
The authors often acknowledge the fact that the study area is mountainous and urbanized. However there is not reference, no discussion and no comparison against studies that include physical processes such as sediment transport and turbidity in the flood modeling as well as concerning the urban drainage network. What the assumptions used, and the hypothesis made by the authors concerning these (neglected) concepts? I believe this should be clarified and offer interesting points to be discusses.
The validation procedure is not clear. Two concepts should be clarified: i) inundation points: the authors should spell out if they are points, cross sections, groups of flooded pixels. Are they flooded by a single or multiple flood events? Are those measured localized data? ii) the authors use the 6cm simulation as benchmark, but no spatial error metrics are presented over the analyzed domain. This is crucial for having an overall model assessment with varying resolution and not only at the 6 points.
Minor comments:
- Figure 1: Inundation points and the corresponding text are not readable from the insert of Figure 1 please use a different color
- Table 1: serial number is not a number but a letter please be consistent in the table description and in the text.
- Line 88: simulation errors: errors or dissimilarities, if errors please clarify how these errors are computed, against what type of ground truth?
- Line 116: please clarify what is the meaning of “typycal” and what are these main feautures
- Line 116: According to the 2022 Flood Control Plan of Xuanhan: I wonder if the domain was already used in the According to the 2022 Flood Control Plan of Xuanhan and if inundation maps are available in the plan or by local authorities over the same study area. This will provide an independent validation of the method used in this paper, which move from 6 points to spatial assessment.
- Line 118: what is a warning points, how is this defined? in a location/a cross section, an area of pixels well defined where inundation was observed? if yes, what about the surrounding of the 6 points?
- Line 121: please be more explicit on how did you select the flood event (if it was only one or more than one); is this measured or simulated? if simulated by which kind of model? how about initial hydrological/hydraulic conditions that generated this event?
- Line 134: can you better define the PCI Geomatica, which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
- Line 137: carried out using the highest quality settings: can you please specify what does it means? which processes are involved in the “high quality“ and how “high“ is defined?
- Line 148: processing tool RAS Mapper: which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
- Line 150: a table could be useful where the authors clearly mention all the input, the parameters for setting the model simulations, and all the output. This should clarify the procedure and eventually the study reproducibility, which is crucial in research.
- Lines 156-162 please clarify the meaning, the units, and values of the variables presented
- Line 162 radius: please space is missing
- Line 173: what is a flood process: do you mean a flood hydrograph?
- Line 181: how and why did you use these attributes? the meaning of each of them is clear but how and why they will be used in the methodology is not clear.
- Line 201-204: this is not results, it is method; still not clear what are the 6 points
- Line 208 210: what physically happen in correspondence of these six points/cross sections? change of slope, drainage directions? roughness?
- Figure 7: what are the dashed lines in panel b? please define them
- Line 224-225 please the sentence does not make sense, consider to revise or remove the .
- Line 225: "The reason for the lack of significant trends " do you mean statistically significant or only a clear pattern? please clarify
- Line 258: remove 1 dot
- Line 266, please define what the numbers in table 3 mean: eventually with a formula. In the caption mean error is used but how is it computed and why average error is considered; can you add more typical measures such as rmse, absolute error, percentage bias?
Citation: https://doi.org/10.5194/egusphere-2024-404-RC2 -
AC2: 'Reply on RC2', Xiaorong Huang, 23 Oct 2024
Author Comments to Referee#2
Responses to the general review comments:
1.“Being the topography one crucial element of the paper there is a need to define once in the text the definition of dsm and dtm, their differences and which one of them is used in the model. The post processing of the dsm should be detailed in order to make it clear and reproducible. The authors should move from spelling the name of the software used to spelling the single steps of the procedure (and the main algorithms) they used. Is there any computation of levees lines, special topographic features/lines, artificial depressions filling?”
Response: Thank you for your insightful comments. Sorry for the confusion in our expression regarding DSM and DTM. In most past studies, DTM (which focuses only on the natural terrain without surface features like trees or buildings) was used as the input for hydraulic models. However, in this study, since it involves flood inundation simulation, we retained the riverside buildings while processing the DSM obtained from the drone (DSM includes all surface objects), and only filtered out noise points from vegetation, water surfaces, and roads. As a result, the terrain data is in a state between DSM and DTM. After discussion and considering other reviewers' suggestions, we have decided that it is more appropriate to define the terrain data we used as DTM. We will supplement the relevant modifications and explanations in the section on processing drone imagery.
The specific processing method of DSM involves using various filters available in PCI Geomatica software, such as Terrain filters, Pit and Bump Filters, Median filters, and Clamp filters, along with manual touch-up edits to filter the designated vegetation areas.The specific process of handling DSM in this study is indeed important for the readers. Previously, we did not provide detailed explanations due to space constraints. We will supplement the method section with the specific formulas used by the filters and clarify which filters were applied in this study.
When using HEC-RAS for hydraulic simulation, we did not set additional levee lines or other special topographic features/lines, because, unlike traditional low-resolution DEMs, the 6 cm resolution DSM can accurately and realistically reflect terrain obstacles (such as levees and highlands). This was also verified during the cross-section processing and subsequent simulations in HEC-RAS. As for the artificial depressions filling, we handled it using Pit and Bump Filters during the DSM processing.
2. “The authors often acknowledge the fact that the study area is mountainous and urbanized. However there is not reference, no discussion and no comparison against studies that include physical processes such as sediment transport and turbidity in the flood modeling as well as concerning the urban drainage network. What the assumptions used, and the hypothesis made by the authors concerning these (neglected) concepts? I believe this should be clarified and offer interesting points to be discusses.”
Response: Thanks for your professional suggestions. We apologize for not providing a more detailed explanation regarding the impact of sediment transport and urban drainage networks in the manuscript. The reason we did not consider the influence of sediment transport in this study is due to the fact that the research area is located immediately downstream of a large reservoir, which intercepts sediment and results in the release of clear water. Additionally, we referred to the situation in many flood-prone mountainous riverside cities, where—except for extremely underdeveloped areas—reservoirs or other water-retaining structures are typically built upstream to mitigate the effects of flood disasters. These structures also reduce the downstream impact of sediment transport, even during flood discharges, as they provide a degree of sediment interception. Furthermore, since our research focuses on the main urban river section most affected by floods, without considering the broader flood propagation, we concluded that sediment transport has a limited impact on the flood simulation in our study.
As for the impact of urban drainage networks, as mentioned in the introduction of the manuscript, unlike plain cities where urban flooding is mainly caused by impervious surfaces and drainage networks, riverside cities in mountainous areas have impervious roads and various buildings constructed along significantly sloped terrains. During heavy rainfall, runoff rapidly converges towards the lowest areas of the river channel, making it difficult for significant urban flooding to occur. Additionally, flash floods in such areas are dominated by fluvial flood, with the inundation range gradually spreading along both sides of the river. The main human and economic damages occur close to the riverbanks. The rapid rise and fall characteristics of flash floods mean that the processes of flood initiation, peak passage, and recession could all be completed before the urban drainage network in mountainous cities has any significant impact. The main factors influencing flood propagation in such areas are the downstream channel’s flow capacity and upstream inflow variations, with the influence of the urban drainage network being relatively minor in comparison to these factors.
We will add the corresponding explanation to the Materials and Methods section, and we would like to once again thank you for pointing out these key issues.
3.“The validation procedure is not clear. Two concepts should be clarified: i) inundation points: the authors should spell out if they are points, cross sections, groups of flooded pixels. Are they flooded by a single or multiple flood events? Are those measured localized data? ii) the authors use the 6cm simulation as benchmark, but no spatial error metrics are presented over the analyzed domain. This is crucial for having an overall model assessment with varying resolution and not only at the 6 points.”
Response: Thank you for raising these important points. We sincerely apologize for the lack of clarity in the validation procedure. In fact, the term "inundation point" refers to a specific location indicating the flood boundary, which in reality is a flood boundary line. This line represents the position that the flood reaches when the water flow attains a certain value. These six flood boundary locations were selected by the local flood management department as key flood warning and observation points. During the flood season, staff members are stationed at these points to observe conditions and oversee evacuation procedures.
The flood boundary lines for each location were derived from historical observations over multiple flood events, which are summarized in the form of red lines (as shown in the field investigation photos in Figure 5). This data was provided by the technical staff of the Dazhou Hydrological and Water Resources Survey Centre (one of the co-authors of this study). During our field investigation, we also verified these lines through the inspection of historical flood traces and interviews with local residents.We will include this additional explanation in the relevant sections of the manuscript.
Regarding the issue of missing spatial error metrics, we fully agree with the reviewer’s opinion. For the entire study area, we only used inundation area and average inundation depth as evaluation metrics, which are indeed insufficient for a comprehensive assessment of the model. Moving forward, we plan to include additional metrics to statistically evaluate the differences in flood depth and inundation extent simulations.
Responses to the minor comments:
1.Figure 1: Inundation points and the corresponding text are not readable from the insert of Figure 1 please use a different color
Response:We appreciate the reviewer for pointing out this issue. We will revise the colors of the inundation points in Figure 1 accordingly.
2.Table 1: serial number is not a number but a letter please be consistent in the table description and in the text.
Response: Thank for your suggestion. We will correct the erroneous statements in Table 1.
3.Line 88: simulation errors: errors or dissimilarities, if errors please clarify how these errors are computed, against what type of ground truth?
Response: Thanks for your comment. We will supplement the calculation methods for the errors and the types of ground truth used for comparison, along with the corresponding references.
4.Line 116: please clarify what is the meaning of “typycal” and what are these main feautures
Response:Thanks for your comment. We will provide additional explanations for the features of typical mountainous riverside cities, including Geographical Location, Topography, and Urban Distribution.
5.Line 116: According to the 2022 Flood Control Plan of Xuanhan: I wonder if the domain was already used in the According to the 2022 Flood Control Plan of Xuanhan and if inundation maps are available in the plan or by local authorities over the same study area. This will provide an independent validation of the method used in this paper, which move from 6 points to spatial assessment.
Response:Thanks for your comment.Relevant explanations can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
6.Line 118: what is a warning points, how is this defined? in a location/a cross section, an area of pixels well defined where inundation was observed? if yes, what about the surrounding of the 6 points?
Response:Thanks for your comment.Relevant explanations can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
7.Line 121: please be more explicit on how did you select the flood event (if it was only one or more than one); is this measured or simulated? if simulated by which kind of model? how about initial hydrological/hydraulic conditions that generated this event?
Response:Thanks for your comment. We used observational data from the major flood event in 2005, and we will revise the unclear expression.
8.Line 134: can you better define the PCI Geomatica, which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
Response:Thanks for your comment.Relevant explanations can be found in the response to the first major comment above, and we will also include additional clarifications in the manuscript.
9.Line 137: carried out using the highest quality settings: can you please specify what does it means? which processes are involved in the “high quality“ and how “high“ is defined?
Response:Thanks for your comment.The term "high quality" refers to an option in DJI's drone imaging processing software, DJI Terra. High-quality processing signifies the generation of denser and more refined point cloud data, utilizing advanced noise filtering techniques, among others. This approach places higher demands on both image quality and computational hardware. We will provide corresponding explanations in the manuscript.
10.Line 148: processing tool RAS Mapper: which tools algorithm did you apply for obtain the filtered maps? Can you describe them in order to achieve reproducible research?
Response:Thanks for your comment.One of the major problems in hydraulic modeling is that terrain data does not often include the actual terrain underneath the water surface in the channel region. RAS Mapper can now be used to create a terrain model of the channel region from the HECRAS cross sections and the Cross Section Interpolation Surface. This terrain model can then be combined with the general surface terrain model (that does not accurately depict the terrain below the water surface) to create an improved terrain model for hydraulic modeling and mapping.
11.Line 150: a table could be useful where the authors clearly mention all the input, the parameters for setting the model simulations, and all the output. This should clarify the procedure and eventually the study reproducibility, which is crucial in research.
Response:Thanks for your comment.We will include a table in the manuscript that contains all the inputs, the parameters used for setting up the model simulations, and all the outputs.
12.Lines 156-162 please clarify the meaning, the units, and values of the variables presented
Response:Thanks for your comment.We will supplement the meaning, units, and values of the variables presented.
13.Line 162 radius: please space is missing
Response:Thanks for your comment.We will correct the missing spaces.
14.Line 173: what is a flood process: do you mean a flood hydrograph?
Response:Thanks for your comment.This refers to using the runoff time series of the selected flood events as input. We will revise the wording that caused the confusion.
15.Line 181: how and why did you use these attributes? the meaning of each of them is clear
but how and why they will be used in the methodology is not clear.
Response:Thanks for your comment.These six attributes comprehensively account for the factors of elevation, terrain ruggedness, exposure, and morphometric protection, as well as flow resistance, providing a thorough evaluation of how topographic features influence hydrodynamics. For instance:
Elevation: Elevation is a critical topographic parameter affecting water flow speed, direction, and energy. It directly determines the downhill flow tendency, making it essential for flood simulation.
TPI (Topographic Position Index): TPI describes the position of a point within its surrounding terrain (e.g., hilltops or valleys). It helps determine areas where water collects or disperses, influencing flow paths and ponding locations.
TRI (Terrain Ruggedness Index): TRI measures terrain roughness, describing the degree of surface undulation. Rugged terrain can increase flow friction, affecting both flow velocity and water kinetic energy.
MPI (Morphometric Protection Index): MPI reflects the protective role of terrain against natural disasters like landslides or erosion. In hydraulic models, this index helps assess the influence of terrain on flow paths.
VRM (Vector Ruggedness Measure): VRM is another measure of terrain complexity, capturing surface variation in all directions. Complex terrain can significantly alter water flow paths and velocities, and in flood simulations, VRM assists in identifying areas where water flow might be obstructed.
We will incorporate this clarification into the Table 2.
16.Line 201-204: this is not results, it is method; still not clear what are the 6 points
Response:Thanks for your comment. We will supplement and expand the content of this section, and the explanation regarding the inundation points can be found in the response to the third major comment above, and we will also include additional clarifications in the manuscript.
17.Line 208 210: what physically happen in correspondence of these six points/cross sections? change of slope, drainage directions? roughness?
Response:Thanks for your comment. The explanation regarding the inundation points can be found in the response to the third major comment above.
18.Figure 7: what are the dashed lines in panel b? please define them
Response:Thanks for your comment.The dashed lines are used to better illustrate the variations (step changes) between the simulation results at different resolutions. We will add relevant explanations in the manuscript.
19.Line 224-225 please the sentence does not make sense, consider to revise or remove the .
Response:Thanks for your comment.We will refine the expression.
20.Line 225: "The reason for the lack of significant trends " do you mean statistically significant or only a clear pattern? please clarify
Response:Thanks for your comment.The term "not significant" here refers to the observation of a clear pattern derived solely from the bar charts in the figure, as it only involves results from six different resolutions (which provides a limited statistical sample), and there are no apparent differences in the results.
21.Line 258: remove 1 dot
Response:Thanks for your comment.We will correct this error.
22 .Line 266, please define what the numbers in table 3 mean: eventually with a formula. In the caption mean error is used but how is it computed and why average error is considered; can you add more typical measures such as RMSE, absolute error, percentage bias?
Response:Thanks for your professional suggestions. We will revise the results in Table 3 and their corresponding descriptions, and we will also add other typical measures.
Thank you very much for taking the time to review our manuscript and for providing many professional evaluations. Each comment has been extremely helpful in enhancing the professionalism and readability of the manuscript. We will carefully revise the manuscript according to your suggestions. Once again, we appreciate the valuable feedback provided by the reviewer.
Citation: https://doi.org/10.5194/egusphere-2024-404-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 339 | 157 | 311 | 807 | 21 | 22 |
- HTML: 339
- PDF: 157
- XML: 311
- Total: 807
- BibTeX: 21
- EndNote: 22
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 249 | 29 |
| China | 2 | 84 | 9 |
| Germany | 3 | 47 | 5 |
| Romania | 4 | 35 | 4 |
| Denmark | 5 | 31 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 249