the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 May 2024
A multiscale modelling framework of coastal flooding events for global to local flood hazard assessments
Abstract. Tropical and extratropical cyclones, which can cause coastal flooding, are among the most devastating natural hazards. Understanding better coastal flood risk can help to reduce their potential impacts. Global flood models play a key role in this process. In recent years, global models and methods for flood hazard simulation have improved, but they still present limitations to provide actionable information at local scales. In order to address some of those limitations we present MOSAIC, a novel modelling framework that couples dynamic water level and overland flood models. MOSAIC follows a multiscale modelling approach in which local models with high-resolution are nested within a coarser large-scale model to obtain higher-resolution water levels and provide better coastal boundary conditions for dynamic flood modelling. To demonstrate the capabilities of MOSAIC we simulate three historical storm events. To merit the potential of MOSAIC’s multiscale modelling approach we perform a sensitivity analysis. Our findings indicate that various model refinements influence the simulation of total water levels and flood depths. The degree of importance of each refinement is linked to the local topography of the study area, the spatial heterogeneity of the water levels and the storm characteristics. MOSAIC provides a bridge between fully global and fully local modelling approaches, paving the way towards more actionable large-scale flood risk assessments.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Natural Hazards and Earth System Sciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(3207 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3207 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1354', Anonymous Referee #1, 03 Aug 2024
Dear Authors,
After reviewing your manuscript, I find that it holds significant relevance and potential. However, in its current form, it fails to address any substantial scientific question. The results presented are merely a model-model comparison, illustrating that variations in spatial or temporal resolution affect the outcomes. This does not address the scientific questions or limitations introduced in the introduction. Therefore, I recommend major revisions and provide the following comments to help strengthen the manuscript and enhance its scientific rigor.
1. Introduction: This section is well-written but could benefit from more specifics regarding the dynamic processes that are currently missing (L46-51). For instance, details on wave-driven processes, hydrological processes, and man-made structures would be valuable. Additionally, I challenge the notion that the limitation of topo-bathymetry in global applications can be resolved solely through grid refinement. In my mind, there are three main methodological challenges: resolution, input data sets (topo-bathy and others), and physical processes. This paper addresses the first one but not the other two. Hence, the linkage from the scientific gap to the approach does not hold, as the MOSAIC modeling framework does not resolve challenges with input data nor does it address additional processes relevant for inundation that global models fail to account for.
2. MOSAIC Modeling Framework: The authors aim to introduce a modeling framework. To do so successfully, a more comprehensive introduction to other modeling frameworks and/or nesting techniques is necessary. Additionally, more details are needed on what has been specifically programmed and what is novel about it. For example, details on the Holland parametric wind model and how it is integrated are missing. I also miss details on the nesting procedure used for both the offline Delft3D FM and SFINCS approach. A more rigorous description of the code would enhance the scientific value of the manuscript.
3. Modeling Results: This section requires the most work. As mentioned, in its current state, it is a model-model comparison without any significant insights. To address this, I strongly recommend the authors include relevant observations of observed water levels and flood extents. Without these, it is difficult to assess differences and model accuracy. The insights regarding the relevance of temporal and spatial resolution require more simulations to assess convergence. For example, the authors could show differences between 1, 2, 5, 10, 20, and 60 minutes of temporal resolution to demonstrate how water levels respond to these changes. A similar approach can be taken for spatial resolution. During these comparisons, please avoid varying the bathymetry source simultaneously, as this would complicate the findings. When performing these analyses, provide an analysis that supports the findings. Why are water levels higher or lower with these settings?
I am particularly skeptical about the dynamic downscaling/fully refined results. How do the authors explain a 40 cm increase in water level? It seems there might be a double-counting of the inverse barometer effect or another error. I do not believe that the entire Gulf of Mexico can have such a different water level based on minor model configuration changes. Could the authors provide more justification for these findings? To understand the results better, I recommend analyzing the time series first.In this section, the authors use the word 'might' frequently. I suggest analyzing the results to test these hypotheses. For example, why are the results different for Haiyan with a 60-minute temporal resolution? One can demonstrate this by comparing water levels near the eye of the storm and further away, providing results rather than hypotheses.
In the flood section, the results are unconvincing. For example, during Irma, Jacksonville experienced severe flooding. In Figure 7 (a-d), the city appears unaffected. This is problematic. I suspect that topo-bathymetry is the cause, which brings us back to the challenges mentioned in the introduction that MOSAIC does not resolve. Demonstrating 1) accurate water levels near the city and 2) flood extents with more reliable US-based topo-bathymetry are essential to successfully model this case study.
4. Discussion: I could not find the MOSAIC code on Zenodo, so I argue that this needs to be shared first before claiming it is 'automated and reproducible.' I also challenge the statement "enhance the simulation at the local scale by providing refined water levels." I have not seen evidence of this in the manuscript.
I hope these comments help improve the manuscript and make it more scientifically robust.Citation: https://doi.org/10.5194/egusphere-2024-1354-RC1 -
AC1: 'Reply on RC1', Irene Benito Lazaro, 11 Oct 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1354/egusphere-2024-1354-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Irene Benito Lazaro, 11 Oct 2024
-
RC2: 'Comment on egusphere-2024-1354', Anonymous Referee #2, 07 Aug 2024
This manuscript presents a methodology for evaluating flooding at high resolution by coupling three models: GTSM, Delft3D, and SFINCs. These models are among the latest and most robust developments in hydrodynamic modeling. On one hand, it is necessary to develop robust methodologies to assess coastal flooding, taking into account different types of forcings such as tropical cyclones (TCs) and extratropical cyclones (ETCs). On the other hand, this study tries to emphasize the importance of increasing both temporal and spatial resolution as well as enhancing hydrodynamic flood modeling.
Therefore, this research and development could be a valuable contribution to the scientific community. However, I believe this study fails to convincingly demonstrate that refining, downscaling, and dynamic flood modeling significantly improve flood hazard assessment. Several aspects need clarification, and a clear message about your results, conclusions, or recommendations for performing a flood risk assessment has not been adequately addressed.
The weakest part of your study is the sensitivity analysis of the model configurations.
First, concerning the organization, it would be advisable to assign a nomenclature to each configuration to aid in comparisons and analysis. For example, the default configuration and the refined temporal and spatial output could be assigned the same letter with different numbering (since they all originate from the same model, with the same forcing and bathymetry). The fully refined configuration should have another letter, because although it combines higher temporal and spatial resolution, the fact that the GTSM simulations use more current and detailed bathymetry (despite Europe EDMOnet being used with the same resolution as GEBCO2014) introduces another distinct element that can significantly affect the results. Finally, the dynamic downscaling (a nomenclature related to the global configuration, to which the fully refined configuration is nested), should not be directly compared with the default configuration without analyzing the effect of the previous factors.
This distinction between configurations should also be used for comparing results at the storm surge modeling and hydrodynamic modeling levels. This way, the analyses would be more orderly, identifying how each factor influences the outcomes, leading to more conclusive comparisons and differences.
In section 3.1 (multiscale storm surge modeling), I would divide it into three subsections: the first analyzing the effect of higher resolution on the maximum water level value (results shown in Figure 5 b, e, and h), the second identifying the effect of bathymetry changes (moving from GEBCO2014 to GEBCO2023, results shown in Figure 4), once the time resolution effect is identified, and the third extracting the added value of dynamic downscaling (once the effects of time resolution and bathymetry changes are understood).
In section 3.2 (Hydrodynamic flood modeling), I would also make a stepwise comparison, isolating the effect of different factors (spatial resolution, temporal resolution, bathymetry, and dynamical downscaling). In the first subsection, compare the default configuration with the refined temporal resolution and refined spatial scale. In the second subsection, compare the results of the fully refined configuration with the dynamical downscaling to isolate and identify this effect. The goal is to see what each factor contributes and how high-resolution modeling with SFINCs improves, once the effect of water level (as the boundary condition) is accounted for. Comparing the default configuration with the dynamical downscaling does not reveal which factor is more influential, for example, if a more detailed bathymetry in the global model is already the most determining factor.
It would be beneficial to conclude with some recommendations, as this methodology can be applied by various researchers and consultants in their flooding studies. It should be noted that the available water level data are related to the default configuration. Therefore, it is crucial to identify the added value of each factor rather than jumping directly to the dynamical downscaling nested in the refined spatial output. Other researchers applying this methodology would start from the water level of the default configuration, even without modeling the TC forced by the Holland wind model (as in the Global Sea Level time series available at Copernicus Climate Data Store: https://doi.org/10.24381/cds.a6d42d60). How well or how it may affect using the Holland model?
Another significant drawback of your study is the lack of validation with observations of water levels and flood extents. Is there any possibility to validate the steps of the MOSAIC methodology with observations of water levels or spatial flooding maps? How can we be sure that increasing spatial/temporal resolution, improving bathymetry, and employing dynamic downscaling enhance the results?
The effects of waves (especially infragravity energy), precipitation, and river discharge should be addressed and discussed more thoroughly in the introduction (other methodologies that take into account these factors) and in the discussion (what effects could be missed) .
Specific question:
Local high resolution model domain: why these domains have been selected? Have they been defined based on the cyclone tracks?
Minor comments:
GEBCO 2020 in line 119 while GEBCO2023 is mentioned in Table 1.
Realistic configuration in Figure 8: not sure what result/output is this.
Citation: https://doi.org/10.5194/egusphere-2024-1354-RC2 -
AC2: 'Reply on RC2', Irene Benito Lazaro, 11 Oct 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1354/egusphere-2024-1354-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Irene Benito Lazaro, 11 Oct 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1354', Anonymous Referee #1, 03 Aug 2024
Dear Authors,
After reviewing your manuscript, I find that it holds significant relevance and potential. However, in its current form, it fails to address any substantial scientific question. The results presented are merely a model-model comparison, illustrating that variations in spatial or temporal resolution affect the outcomes. This does not address the scientific questions or limitations introduced in the introduction. Therefore, I recommend major revisions and provide the following comments to help strengthen the manuscript and enhance its scientific rigor.
1. Introduction: This section is well-written but could benefit from more specifics regarding the dynamic processes that are currently missing (L46-51). For instance, details on wave-driven processes, hydrological processes, and man-made structures would be valuable. Additionally, I challenge the notion that the limitation of topo-bathymetry in global applications can be resolved solely through grid refinement. In my mind, there are three main methodological challenges: resolution, input data sets (topo-bathy and others), and physical processes. This paper addresses the first one but not the other two. Hence, the linkage from the scientific gap to the approach does not hold, as the MOSAIC modeling framework does not resolve challenges with input data nor does it address additional processes relevant for inundation that global models fail to account for.
2. MOSAIC Modeling Framework: The authors aim to introduce a modeling framework. To do so successfully, a more comprehensive introduction to other modeling frameworks and/or nesting techniques is necessary. Additionally, more details are needed on what has been specifically programmed and what is novel about it. For example, details on the Holland parametric wind model and how it is integrated are missing. I also miss details on the nesting procedure used for both the offline Delft3D FM and SFINCS approach. A more rigorous description of the code would enhance the scientific value of the manuscript.
3. Modeling Results: This section requires the most work. As mentioned, in its current state, it is a model-model comparison without any significant insights. To address this, I strongly recommend the authors include relevant observations of observed water levels and flood extents. Without these, it is difficult to assess differences and model accuracy. The insights regarding the relevance of temporal and spatial resolution require more simulations to assess convergence. For example, the authors could show differences between 1, 2, 5, 10, 20, and 60 minutes of temporal resolution to demonstrate how water levels respond to these changes. A similar approach can be taken for spatial resolution. During these comparisons, please avoid varying the bathymetry source simultaneously, as this would complicate the findings. When performing these analyses, provide an analysis that supports the findings. Why are water levels higher or lower with these settings?
I am particularly skeptical about the dynamic downscaling/fully refined results. How do the authors explain a 40 cm increase in water level? It seems there might be a double-counting of the inverse barometer effect or another error. I do not believe that the entire Gulf of Mexico can have such a different water level based on minor model configuration changes. Could the authors provide more justification for these findings? To understand the results better, I recommend analyzing the time series first.In this section, the authors use the word 'might' frequently. I suggest analyzing the results to test these hypotheses. For example, why are the results different for Haiyan with a 60-minute temporal resolution? One can demonstrate this by comparing water levels near the eye of the storm and further away, providing results rather than hypotheses.
In the flood section, the results are unconvincing. For example, during Irma, Jacksonville experienced severe flooding. In Figure 7 (a-d), the city appears unaffected. This is problematic. I suspect that topo-bathymetry is the cause, which brings us back to the challenges mentioned in the introduction that MOSAIC does not resolve. Demonstrating 1) accurate water levels near the city and 2) flood extents with more reliable US-based topo-bathymetry are essential to successfully model this case study.
4. Discussion: I could not find the MOSAIC code on Zenodo, so I argue that this needs to be shared first before claiming it is 'automated and reproducible.' I also challenge the statement "enhance the simulation at the local scale by providing refined water levels." I have not seen evidence of this in the manuscript.
I hope these comments help improve the manuscript and make it more scientifically robust.Citation: https://doi.org/10.5194/egusphere-2024-1354-RC1 -
AC1: 'Reply on RC1', Irene Benito Lazaro, 11 Oct 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1354/egusphere-2024-1354-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Irene Benito Lazaro, 11 Oct 2024
-
RC2: 'Comment on egusphere-2024-1354', Anonymous Referee #2, 07 Aug 2024
This manuscript presents a methodology for evaluating flooding at high resolution by coupling three models: GTSM, Delft3D, and SFINCs. These models are among the latest and most robust developments in hydrodynamic modeling. On one hand, it is necessary to develop robust methodologies to assess coastal flooding, taking into account different types of forcings such as tropical cyclones (TCs) and extratropical cyclones (ETCs). On the other hand, this study tries to emphasize the importance of increasing both temporal and spatial resolution as well as enhancing hydrodynamic flood modeling.
Therefore, this research and development could be a valuable contribution to the scientific community. However, I believe this study fails to convincingly demonstrate that refining, downscaling, and dynamic flood modeling significantly improve flood hazard assessment. Several aspects need clarification, and a clear message about your results, conclusions, or recommendations for performing a flood risk assessment has not been adequately addressed.
The weakest part of your study is the sensitivity analysis of the model configurations.
First, concerning the organization, it would be advisable to assign a nomenclature to each configuration to aid in comparisons and analysis. For example, the default configuration and the refined temporal and spatial output could be assigned the same letter with different numbering (since they all originate from the same model, with the same forcing and bathymetry). The fully refined configuration should have another letter, because although it combines higher temporal and spatial resolution, the fact that the GTSM simulations use more current and detailed bathymetry (despite Europe EDMOnet being used with the same resolution as GEBCO2014) introduces another distinct element that can significantly affect the results. Finally, the dynamic downscaling (a nomenclature related to the global configuration, to which the fully refined configuration is nested), should not be directly compared with the default configuration without analyzing the effect of the previous factors.
This distinction between configurations should also be used for comparing results at the storm surge modeling and hydrodynamic modeling levels. This way, the analyses would be more orderly, identifying how each factor influences the outcomes, leading to more conclusive comparisons and differences.
In section 3.1 (multiscale storm surge modeling), I would divide it into three subsections: the first analyzing the effect of higher resolution on the maximum water level value (results shown in Figure 5 b, e, and h), the second identifying the effect of bathymetry changes (moving from GEBCO2014 to GEBCO2023, results shown in Figure 4), once the time resolution effect is identified, and the third extracting the added value of dynamic downscaling (once the effects of time resolution and bathymetry changes are understood).
In section 3.2 (Hydrodynamic flood modeling), I would also make a stepwise comparison, isolating the effect of different factors (spatial resolution, temporal resolution, bathymetry, and dynamical downscaling). In the first subsection, compare the default configuration with the refined temporal resolution and refined spatial scale. In the second subsection, compare the results of the fully refined configuration with the dynamical downscaling to isolate and identify this effect. The goal is to see what each factor contributes and how high-resolution modeling with SFINCs improves, once the effect of water level (as the boundary condition) is accounted for. Comparing the default configuration with the dynamical downscaling does not reveal which factor is more influential, for example, if a more detailed bathymetry in the global model is already the most determining factor.
It would be beneficial to conclude with some recommendations, as this methodology can be applied by various researchers and consultants in their flooding studies. It should be noted that the available water level data are related to the default configuration. Therefore, it is crucial to identify the added value of each factor rather than jumping directly to the dynamical downscaling nested in the refined spatial output. Other researchers applying this methodology would start from the water level of the default configuration, even without modeling the TC forced by the Holland wind model (as in the Global Sea Level time series available at Copernicus Climate Data Store: https://doi.org/10.24381/cds.a6d42d60). How well or how it may affect using the Holland model?
Another significant drawback of your study is the lack of validation with observations of water levels and flood extents. Is there any possibility to validate the steps of the MOSAIC methodology with observations of water levels or spatial flooding maps? How can we be sure that increasing spatial/temporal resolution, improving bathymetry, and employing dynamic downscaling enhance the results?
The effects of waves (especially infragravity energy), precipitation, and river discharge should be addressed and discussed more thoroughly in the introduction (other methodologies that take into account these factors) and in the discussion (what effects could be missed) .
Specific question:
Local high resolution model domain: why these domains have been selected? Have they been defined based on the cyclone tracks?
Minor comments:
GEBCO 2020 in line 119 while GEBCO2023 is mentioned in Table 1.
Realistic configuration in Figure 8: not sure what result/output is this.
Citation: https://doi.org/10.5194/egusphere-2024-1354-RC2 -
AC2: 'Reply on RC2', Irene Benito Lazaro, 11 Oct 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1354/egusphere-2024-1354-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Irene Benito Lazaro, 11 Oct 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 685 | 232 | 275 | 1,192 | 44 | 50 |
- HTML: 685
- PDF: 232
- XML: 275
- Total: 1,192
- BibTeX: 44
- EndNote: 50
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Jeroen C. J. H. Aerts
Philip J. Ward
Dirk Eilander
Sanne Muis
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3207 KB) - Metadata XML