the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
iHydroSlide3D v1.0: an advanced hydrological-geotechnical model for hydrological simulation and three-dimensional landslide prediction
Guoding Chen
Sheng Wang
Yi Xia
Lijun Chao
Abstract. Forecasting flood–landslide cascading disasters in flood- and landslide-prone regions is an important topic within the scientific community. Existing hydrological-geotechnical models mainly employ infinite or static 3D stability model and very few models have incorporated the 3D landslide model into a distributed hydrological model. In this work, we modified a 3D landslide model to account for slope stability under various soil wetness states and then coupled it with the Coupled Routing and Excess STorage (CREST) distributed hydrology model, forming a new modelling system called iHydroSlide3D v1.0. The model features the feasibility of applying flexibly different simulating resolutions for hydrological and slope stability submodules by embedding a soil moisture downscaling method. For a large-scale application, we paralleled the code and elaborated several computational strategies. The model produces a relatively comprehensive and reliable diagnosis for flood-landslide events, including (i) complete hydrological components (e.g., soil moisture and streamflow), (ii) a landslide susceptibility assessment (factor of safety and probability of occurrence), and (iii) a landslide hazard analysis (geometric properties of potential failures). We evaluated the plausibility of the model by testing it in a large and complex geographical area, the Yuehe River Basin, China, where we attempted to reproduce cascading flood–landslide events. The results are well verified at both hydrological and geotechnical levels. iHydroSlide3D v1.0 is therefore appropriately used as an innovative tool for assessing and predicting cascading flood–landslide events once the model is well calibrated.
Guoding Chen et al.
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2021-283', Anonymous Referee #1, 23 Feb 2022
Review of iHydroSlide3D v1.0: an advanced hydrological-geotechnical model for hydrological simulation and three-dimensional landslide prediction
Overview
The paper presents an algorithm that couples hydrological modelling and 3-D landslide prediction, and gives a good overview of why such a model is needed given what already exists in that field. The introduction provides a good review of the limitations associated with hydrological modelling and landslide prediction separately, and then how these two types of models can be combined to better represent mass movement. The model framework also provides a clear overview of the method and its underpinning equations. I am recommending this paper for publication with minor revisions. These revisions and suggestions for improvement are as follows:
Introduction
- The scientific and technical contribution of the paper and model is clear, but it would be good for the authors to mention the real-world applications of combined hydrological-geotechnical models for land planning and disaster risk management. Although briefly mentioned at the end (P29 L655-656), discussing the importance of this type of model to policy-makers and decision-makers would further underscore its utility.
Model framework
- Section 2.8: this section needs some revision to present what inputs are needed in a more organized manner. The first sentence gives a high-level overview of what inputs datasets are needed, but the section doesn’t give a good idea of what other parameters are needed in association with these datasets. P13 L337-338 then says that hydrological parameters are needed, but with only one example. Later in Section 3 (P14 L370), it says that information about the impervious surface area was calculated from the land cover map. This section would be more useful if it put the required underlying data and parameters up front. The following table is just a suggestion as a starting point:
Input
Datasets
Derived datasets/parameters
Topographic
Digital elevation model
Flow direction
Flow accumulation
Topographic wetness index
Land cover
Land surface cover
Percentage impervious area
Soil
Soil texture
Saturated hydraulic conductivity
Available water capacity
Exponent of the infiltration curve
Results and discussion
- Like the introduction section, it will be good to have more discussion about how the model framework can contribute to land management and disaster risk management.
- P29 L655-656 says: “The produced zones of risk and landslide geometric properties are valuable for disaster prevention and risk management.”
- Can you give some examples of how you see the model framework and code being used in these situations? Would they be used for climate change scenarios? Would they be used for real-time modelling? Would the risk zones be used as guidelines for no-build areas?
Specific comments
- P3 L90: “physically-based”?
- P4 L100-101: Add access dates to URLs
- P10 L276: Why is the minimum value of FS and PF assigned to the cell?
- P11 L280: What is meant by “sufficiently large number of ellipsoids” and how is that determined? P16 Table 3 talks about the parameter set based on landslide inventory, but how is that calculated?
- P11 Figure 4: The white letters and numbers are hard to see, would recommend a black outline to make them stand out.
- P14 L375: Please give an idea of how much time was needed, either for each module (hydrological and geotechnical) or total runtime to get the user an idea of computational efficiency.
- P16 Table 3: Although defaults are recommended later in the text, it would be good to give some ranges/ballpark figures for Ncores, Landslide Density, TotalTile.
- P21 Figure 9: It is difficult to see the orange landslide circles, would recommend maybe black outline hollow circles instead of orange outline.
- P22 Figure 10: Are these the same four moments as in Figure 9? I would recommend putting the timestamps here too.
- P26 Figure 13: It is hard to see the orange circles again, and they may blend in with the orange colours in the legend. Would again recommend black outline hollow circles.
- P29 L596: What was the computational time required and what is considered “acceptable”?
Code testing
I tested the code available on Zenodo. The included manual is great to show a step-by-step of what needs to be done. It was relatively easy to get it up and running, but I ran into some errors in MATLAB because I had not installed some toolboxes (e.g. Mapping Toolbox, Parallel Computing Toolbox, Curve Fitting Toolbox). It would be good if the manual included the list of Toolbox dependencies in case the user has a limited installation of MATLAB.
Once the Toolboxes were installed, I was able to run the code without errors. I wasn’t sure how to interpret the datasets in the Results section, and it would be good if the manual could give a brief rundown on what the results represent. The accompanying text files (Outlet_Results, Outpix_03501000_Results, Outpix_03501000_Results_Statistics) were also empty.
Citation: https://doi.org/10.5194/gmd-2021-283-RC1 -
AC1: 'Reply on RC1', Ke Zhang, 07 Apr 2022
We highly appreciated the comments from this review. Below are our detailed responses to your comments.
- “Introduction: The scientific and technical contribution of the paper and model is clear, but it would be good for the authors to mention the real-world applications of combined hydrological-geotechnical models for land planning and disaster risk management. Although briefly mentioned at the end (P29 L655-656), discussing the importance of this type of model to policy-makers and decision-makers would further underscore its utility.”
Response:
We have mentioned the real-world applications of combined hydrological-geotechnical models for land planning and disaster risk management in the text. However, thank you for your suggestion. We will provide more information and discussion regarding the importance of this type of model to policy-makers and decision-makers in the revised manuscript.
- “Model framework: Section 2.8: this section needs some revision to present what inputs are needed in a more organized manner. The first sentence gives a high-level overview of what inputs datasets are needed, but the section doesn’t give a good idea of what other parameters are needed in association with these datasets. P13 L337-338 then says that hydrological parameters are needed, but with only one example. Later in Section 3 (P14 L370), it says that information about the impervious surface area was calculated from the land cover map. This section would be more useful if it put the required underlying data and parameters up front. The following table is just a suggestion as a starting point.”
Response:
Thank you for this comment. We will provide a new table (see Table 1) to give an overview of inputs datasets needed in iHydroSlide3D v1.0. The model inputs can be summarized into four types: meteorological forcing data, land surface feature data, simulation parameters, and calibration/verification data. More details on the description, value/resolution, and source of the input data will be provided in Section 3.
- “Results and discussion: Like the introduction section, it will be good to have more discussion about how the model framework can contribute to land management and disaster risk management.
P29 L655-656 says: ‘The produced zones of risk and landslide geometric properties are valuable for disaster prevention and risk management.’
Can you give some examples of how you see the model framework and code being used in these situations? Would they be used for climate change scenarios? Would they be used for real-time modelling? Would the risk zones be used as guidelines for no-build areas?”
Response:
We thank the reviewer for the useful comment. We will add a new paragraph to discuss this point. The comprehensive assessments (in both flood and landslide) possibly contribute to land management and disaster risk management with professional analysis. The landslide susceptibility and hazard zoning are able to manage landslide hazard in urban/rural areas by excluding development in higher hazard areas, and requiring hydro-geotechnical assessment in the planning stage (Fell et al., 2008). The conception has been introduced across some countries such as France (Fell et al., 2008) and Switzerland (Leroi et al., 2005). A recent work corroborated existing hypotheses that urbanization increases landslide hazards (Johnston et al., 2021). Our model could be used as a tool to study the importance of considering interactions with urbanization when predicting landslide hazards under climate change scenarios. The current modular framework and flexibility of the modelling setup also make it feasible to link with other numerical weather prediction models and real-time forcings. We would like to stress that these complicated applications generally require extraordinary computing resources to support. Regarding your above questions, we will provide more explanation and discussion in the revised text.
- For all typos and minor comments pointed out by the reviewer
Response:
Many thanks for pointing them out. We will corrected them in the revised manuscript and do a futher grammar checking.
- “P10 L276: Why is the minimum value of and assigned to the cell?”
Response:
For regional modelling, our 3D landslide model randomly generates a large number of ellipsoidal potential landslides. As a result, every single grid cell will participate in the stability analysis of different landslides multiple times and thus brings a contradiction to the value of (see Fig. 4). The model did not provide the minimum but defined it by counting values (see Fig. 4 and Eq. 28). We acknowledge that the model outcomes the minimum value of ( in the manuscript, and corresponds to the worst-case situation). However, we also compute the value to enhance the analysis. This point has been noted in the original manuscript (P12 L305) and will be further clarified in the revised version to avoid confusion.
- “P11 L280: What is meant by “sufficiently large number of ellipsoids” and how is that determined? P16 Table 3 talks about the parameter set based on landslide inventory, but how is that calculated?”
Response:
In this work, a dimensionless value “landslide density” is defined to reflect the number of ellipsoids. We carried out several convergence tests to find the reasonable value for landslide density (Fig. 7). The so-called “sufficiently large number” means that we have gotten a steady (or convergent) result. The detailed analysis has been shown in Sect. 4.2. We will also add more explanation in the revise manuscript.
The parameters set based on landslide inventory in Table 3 are the maximum and minimum of landslide dimensions (length and width). Our inventory did not record the dimensional information for all landslides, but a few of them. We extract the maximum and minimum of length/width from this limited dataset. The max/min values therefore comprised the constraints for landslide modelling. Considering the random interval equals the spatial resolution (12.5 m in this work), the constraint boundaries were rounded to the integer for simplification. We will add more explanation in Section 3.
- “P14 L375: Please give an idea of how much time was needed, either for each module (hydrological and geotechnical) or total runtime to get the user an idea of computational efficiency.”
Response:
It’s not easy to record the runtime for each module because the whole model was executed in a coupled manner. However, we will provide more informaiton on the total runtime.
- “Code testing: I tested the code available on Zenodo. The included manual is great to show a step-by-step of what needs to be done. It was relatively easy to get it up and running, but I ran into some errors in MATLAB because I had not installed some toolboxes (e.g. Mapping Toolbox, Parallel Computing Toolbox, Curve Fitting Toolbox). It would be good if the manual included the list of Toolbox dependencies in case the user has a limited installation of MATLAB.
Once the Toolboxes were installed, I was able to run the code without errors. I wasn’t sure how to interpret the datasets in the Results section, and it would be good if the manual could give a brief rundown on what the results represent. The accompanying text files (Outlet_Results, Outpix_03501000_Results, Outpix_03501000_Results_Statistics) were also empty.”
Response:
We thank the helpful feedback. The code does rely on some toolboxes of MATLAB. We have improved the manual by addressing this point. In addition to the most basic installation content, the software requires additional toolboxes: Curve Fitting Toolbox, Global Optimization Toolbox, Optimization Toolbox, Partial Differential Equation Toolbox, Statistics and Machine Learning Toolbox, Symbolic Math Toolbox, MATLAB in the Cloud, MATLAB Parallel Server, and Parallel Computing Toolbox. For users who have most of the toolbox installed by default, they can run the software easily. For users who have a limited installation of MATLAB, they are advised to check the availability of toolboxes in advance. Users can also directly run the code without concerns and install the toolboxes needed according to the error prompt.
We have added the explanation for all possible outputs in our updated manual. The empty text file is a small bug with csv file writing and we have fixed it in our updated version of iHydroSlide3D v1.0. The model provides results for the outlet location, which are written to csv format series file. The model will automatically calculate the output pixel and output the csv file named “Outlet_Results”. Users may also choose any specified output pixels so long as they are within the given watershed extent. For this purpose, users can define it in “ControlFile”. Our example chose the pixel location the same as basin outlet for example (named “Yuehe”). As a consequence, the result file for the output pixel location will be named "Outpix_Yuehe_Results". If the observed streamflow is provided in OBS fold, three statistical metrics Nash–Sutcliffe coefficient of efficiency (NSCE), Pearson correlation coefficient (CC), and relative bias will be computed and stored in "Outpix_Yuehe_Results_Statistics". Please note that abnormalities may exist in statistics if streamflow data are missing or all zeros, for which we would like to suggest a simulation including a complete flooding event. The above information is also detailedly described in the updated manual. We will also sync the code on GitHub to Zenodo this discussion is closed.
Citation: https://doi.org/10.5194/gmd-2021-283-AC1
Guoding Chen et al.
Guoding Chen et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,287 | 190 | 34 | 1,511 | 12 | 8 |
- HTML: 1,287
- PDF: 190
- XML: 34
- Total: 1,511
- BibTeX: 12
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1