the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Quasi RealTime Tsunami Early Warning Assessment Based on Finite Fault models
Abstract. In Chile, Tsunami Early Warning relies on a precomputed set of scenarios for places near the tsunami generation zone. These scenarios are nonrealistic and are computed from a uniform constant slip over the rupture surface, because of the short reaction time for the local authorities to manage the threat. Here, we present a new methodology that allows quasireal time tsunami modelling using a finite fault model inversion to evaluate tsunami threat levels. A linear approximation at first order terms of the shallow water wave equations turns into a numerical solver that can be implemented on a programming language. As a case of study, the proposed method is applied in Chile. The results show that it is possible to obtain realistic threat levels and arrival times for the tsunami in progress. Once the finite fault is calculated, it takes a minute to produce the warning maps.
The proposal considers realistic heterogeneous and kinematic fault models of seismic sources obtained rapidly using continuous GPS, strong motion and broadband records, expanding the evaluation capabilities of Tsunami Early Warning Systems. Using directly the finite fault avoids the unrealistic model of uniform slip distribution and diminishes the uncertainty imposed by precomputed scenarios.
 Preprint
(2317 KB)  Metadata XML
 BibTeX
 EndNote
Status: final response (author comments only)

RC1: 'Comment on egusphere20233061', Anonymous Referee #1, 08 Apr 2024
The This is a meaningful study, but its significance lies mainly in the practical tsunami warning operation rather than the science. As a reviewer, I believe that the research has limited scientific innovation as it does not propose its own theories or methods. The methods used in the paper, whether Okada's model or the nonhomogeneous linear shallow water equation, are already mature research outcomes. The primary work of the authors involves discretizing these equations, writing them into programs, and applying them to practical operations.
This work is important, but similar models have been proposed by operational departments in various countries. The authors should at least elucidate the advantages of their method compared to others, such as the one adopted by PMEL, in order to distinguish their approach.
At the same time, I am also curious whether this method has the capability for adjustments during its operation. As shown in Figure 5, the method exhibits insufficient fitting for some stations at the first peak. Can existing models be corrected through observational data?
Overall, I believe that this manuscript is not suitable for publication in NHESS. Unfortunately, I would like to reject this manuscript. Please consider submitting to a more suitable journal. The following comments are for the authors' reference, hoping they can further refine this study.
Line 25: Please change “y” to “and”. As a manuscript submitted to an international journal, it is important to use English expression.
Line 25: The reviewer believes that using the term “sensitive” is not appropriate. Chile is relatively close to the subduction zone, but other countries (such as Japan and Peru) also have similar situations, which cannot be called “sensitive”. Additionally, some areas may face tsunamis triggered by landslides, with similarly short reaction times.
Line 35: Why does the system only consider subduction type earthquakes? Are there any technical issues in considering other types of earthquakes?
Line 50: Although one can understand the meaning of each physical quantity after reading the formulas, I suggest the author explain the meaning of each physical quantity to enhance the readability of the article.
Line 70: The assumption of null time derivative/velocity is based on what evidence? Please specify.
Line 96: What is the scale of the damping layer? How many grid points does it occupy, and has sensitivity testing been conducted?
Line 138: Why is Coriolis force not considered? I understand that the author is mainly calculating nearfield tsunami, but if considering propagation over slightly longer distances, such as Easter Island, Coriolis force needs to be taken into account.
Figure 5: The figure title mentions four events, but please indicate them in the figure.
Citation: https://doi.org/10.5194/egusphere20233061RC1 
AC3: 'Reply on RC1', Matias Sifon, 02 Jun 2024
The This is a meaningful study, but its significance lies mainly in the practical tsunami warning operation rather than the science. As a reviewer, I believe that the research has limited scientific innovation as it does not propose its own theories or methods. The methods used in the paper, whether Okada's model or the nonhomogeneous linear shallow water equation, are already mature research outcomes. The primary work of the authors involves discretizing these equations, writing them into programs, and applying them to practical operations.
This work is important, but similar models have been proposed by operational departments in various countries. The authors should at least elucidate the advantages of their method compared to others, such as the one adopted by PMEL, in order to distinguish their approach.
At the same time, I am also curious whether this method has the capability for adjustments during its operation. As shown in Figure 5, the method exhibits insufficient fitting for some stations at the first peak. Can existing models be corrected through observational data?.
It can be corrected as soon as more data comes in. However the scope of the paper is to enhance the tsunami early warning system. We trade off accuracy for speed. Then even this is possible, the idea is to model the tsunami before the first arrival. So, if the data allows the seismological agency to model the earthquake source before that time, it can be corrected, if not, we get the model with the information available at that time.
Overall, I believe that this manuscript is not suitable for publication in NHESS. Unfortunately, I would like to reject this manuscript. Please consider submitting to a more suitable journal. The following comments are for the authors' reference, hoping they can further refine this study.
Line 25: Please change “y” to “and”. As a manuscript submitted to an international journal, it is important to use English expression.
This is a typo, done.
Line 25: The reviewer believes that using the term “sensitive” is not appropriate. Chile is relatively close to the subduction zone, but other countries (such as Japan and Peru) also have similar situations, which cannot be called “sensitive”. Additionally, some areas may face tsunamis triggered by landslides, with similarly short reaction times.
Thanks for the suggestion, done. (highly vulnerable to tsunami threats)
Line 35: Why does the system only consider subduction type earthquakes? Are there any technical issues in considering other types of earthquakes?
There are no limitations. This can be done for any type of earthquake, but since we focus on Chile, and there are not normal or strike slip events that have generated tsunamis records on sea level instrumentation, we used the ones that have data, which are subduction type events.
Line 50: Although one can understand the meaning of each physical quantity after reading the formulas, I suggest the author explain the meaning of each physical quantity to enhance the readability of the article.
We have explain each term of the equation in the revised manuscript
Line 70: The assumption of null time derivative/velocity is based on what evidence? Please specify.
In the standard modeling. The principle of causality nothing happens before that. Typical modeling.
Line 96: What is the scale of the damping layer? How many grid points does it occupy, and has sensitivity testing been conducted?
We just added a paragraph discussing this issue.
Line 138: Why is Coriolis force not considered? I understand that the author is mainly calculating nearfield tsunami, but if considering propagation over slightly longer distances, such as Easter Island, Coriolis force needs to be taken into account.
We are focused in the near field tsunami modeling. Eastern Island or such locations can be modeled properly with all the terms because this is not near field and the arrival time is not a requirement. The tsunami arrival time could be several hours. A clarification was added.
Figure 5: The figure title mentions four events, but please indicate them in the figure.
This was included in the revised manuscript. The figured has colors for each event.
Citation: https://doi.org/10.5194/egusphere20233061AC3

AC3: 'Reply on RC1', Matias Sifon, 02 Jun 2024

RC2: 'Comments on egusphere20233061', Anonymous Referee #2, 24 Apr 2024
The main concern deals with the use of finite fault models without uncertainty assessment. It is well known that uncertainty sources are intrinsic, and the earthquake and tsunami characteristics remain elusive even several years after the event. In fact, Cienfuegos et al (2019) analyzed 19 different finite fault models of the 2010 Chile earthquake and found a large variability in both coseismic deformation and tsunami intensity measures. Therefore, the use of a single source model solution may not be appropriate. On that ground, the work of Selva et al (2021) proposed a Probabilistic Tsunami Forecasting for tsunami early warning. On the other hand, the manuscript criticized the current Tsunami Warning System in Chile due to the fact that it uses precomputed events of uniform slip over a rupture zone. However, the authors did not mention that the maximum tsunami wave amplitude is selected at each coastal site from a set of scenarios compatible with the ongoing event, which also considers all scenarios within half fault length from the epicenter and with magnitude best approximating the available magnitude incremented by 0.5. Subsequently, despite the current warning system uses uniform slip models, a proxy of existing uncertainty is derived from the statistics of the envelope of selected scenarios. Without uncertainty assessment, the authors assumed that their finite fault model is the correct and final model. However, due to epistemic and aleatory uncertainties there are many sources of error such as preliminary magnitude, epicenter location and slip distribution. I am afraid that the proposed methodology is not an improvement to the existing warning system.
Another major concern is about the novelty of the paper. In addition, it is hard to follow the main idea. Section 2 described (in several pages) the numerical model, and section 3 described the validation of the numerical model. However, if I understood well, the model was previously proposed by Fuentes et al (2019). So, is it necessary to describe the model again here? Or, is this model a new model proposed as part of the methodology? It is not clear in the introduction nor the methodology.
Another concern deals with the poor literature review. Only 18 papers are cited. The introduction does not present the state of the art of tsunami warning systems. With a more comprehensive literature review, the reader could better understand the importance of the proposed methodology. Other tsunami warning systems may be described and not only the Chilean case, such as the discussion can clearly state the advantage and disadvantage of the proposed methodology (What about the warning systems in Italy, Indonesia, US, Japan, etc). Despite the study is focused on the Chilean case, the tentative findings could perfectly be applied to other regions. In addition, important papers are missed such as Cienfuegos et al (2019), Selva et al (2021), Behrens et al (2021).
Another concern deals with the model validation. Lines 228 mentioned that finite fault solutions of 2010, 2011, 2014, 2015 and 2016 earthquakes are the input in the numerical solver. However, it is not mention which finite fault model were used. Are they your own proposed finite fault models? If so, they should be based on the preliminary seismic information to demonstrate that your early tsunami warning system is successful. Or are you just using other authors model to validate the numerical model?. If so, I am confused about the focus of your paper. Are you just proposing a numerical model or are you improving the warning system?
The quality of the paper and written English is very poor. It is important editing and proofreading before manuscript submission.
Section 2 “Methodology” described only the numerical model for tsunami numerical simulation. What about the finite fault model solution? It should be part of the methodology as well.
In section 3 related to the validation of the model, I strongly recommend using the tsunami record at Coquimbo tide gauge. Why is it avoided?
Line 216 mentioned an elliptical finite fault model, but there is no description of this model in the methodology. How is this model generated?
Line 218 mentioned that “after few minutes the FFM is expected to be computed…” but Figure 4 shows the FFM from USGS. I would expect you to show your own finite fault model computed with available information just few minutes after the earthquake.
Line 229, which finite fault models were used? The 2010 has 19 proposed models. It is necessary to indicate the authors in each model. A table would be acceptable. In addition, an explanation on why each model was selected is necessary.
Figure 5. It is necessary to add Coquimbo and Valparaiso to the validation of the Illpapel event. Why are these locations discarded?
Line 247, “…different parameters were used,…” Which parameter are you talking about.
Figure 7 shows different tsunami hazard levels, but theses levels are not defined previously. Is it related to Table 1? Then, it should be mentioned in a Methodology section.
Line 249, “…while figure 7(f) shows the tsunami threat evaluation for far field…” is it Figure 7(g)?
Line 272, “several points were created…” where are these points located? At a given water depth of at the coastline?
Figure 8 is really confusing. Which is a, b or c?, Why the description mentioned twice c? one is with 5h simulation and the other is with 12 h simulation.
Line 296297, “…the seismic source information…” what is the uncertainty related to those parameters? There is no explanation in the paper related to the finite fault model once the seismic parameters are computed.
Line 332, “we model the fault plane defined by the Wphase moment tensor”, but this is not explained in a methodology section. It is necessary to describe the methodology and the time needed to be computed. What is the uncertainty related to this preliminary fault model?
Line 333, “Once the finite fault model is available…” what is the uncertainty of this model, and how long does it take to compute this fault model?
Line 340, “…uses a point source as input”. It should be explained in a methodology section and the time needed to be computed must be mentioned as well.
Figure 9 should be in the results or a methodology section. If this figure is proposed in Riquelme et al (2021), I do not really understand what the novelty of this paper is.
Citation: https://doi.org/10.5194/egusphere20233061RC2 
AC1: 'Reply on RC2', Matias Sifon, 29 May 2024
The main concern deals with the use of finite fault models without uncertainty assessment. It is well known that uncertainty sources are intrinsic, and the earthquake and tsunami characteristics remain elusive even several years after the event. In fact, Cienfuegos et al (2019) analyzed 19 different finite fault models of the 2010 Chile earthquake and found a large variability in both coseismic deformation and tsunami intensity measures. Therefore, the use of a single source model solution may not be appropriate.
It is also not appropriated to use a point source or a uniform source, because this even neglects the physics of earthquakes. Here we present a complementary system that will consdier the best information available at the time that the tsunami threat message needs to be issued for early warning.
On that ground, the work of Selva et al (2021) proposed a Probabilistic Tsunami Forecasting for tsunami early warning. On the other hand, the manuscript criticized the current Tsunami Warning System in Chile due to the fact that it uses precomputed events of uniform slip over a rupture zone.
This is complementary method that can be used along the current tsunami chilean system. A clarification on this was included, we are not criticizing the current strategy, just stating some gaps and proposing a complementary methodology.
However, the authors did not mention that the maximum tsunami wave amplitude is selected at each coastal site from a set of scenarios compatible with the ongoing event, which also considers all scenarios within half fault length from the epicenter and with magnitude best approximating the available magnitude incremented by 0.5.
Incrementing the magnitude in 0.5 it works. So it’s fine but if you think uncertainty is an issue, to raise the magnitude in 0.5 is not real but also means that you haven raised the times energy of the seismic in 5.5 times, which is also arbitrary or 5.6 10 exp 9 Joules.
Subsequently, despite the current warning system uses uniform slip models, a proxy of existing uncertainty is derived from the statistics of the envelope of selected scenarios. Without uncertainty assessment, the authors assumed that their finite fault model is the correct and final model.
No, this is the best information we have available useful before the first arrival time and for tsunami early warning.
However, due to epistemic and aleatory uncertainties there are many sources of error such as preliminary magnitude, epicenter location and slip distribution. I am afraid that the proposed methodology is not an improvement to the existing warning system.
Another major concern is about the novelty of the paper. In addition, it is hard to follow the main idea. Section 2 described (in several pages) the numerical model, and section 3 described the validation of the numerical model. However, if I understood well, the model was previously proposed by Fuentes et al (2019). So, is it necessary to describe the model again here? Or, is this model a new model proposed as part of the methodology? It is not clear in the introduction nor the methodology.
The model is optimized. It has improvements in border conditions in the coast. The parameters used here are fitted for tsunami early warning. These parameters were determined by performance tests. We perform 270 combinations 3 times each, which means 810 tests. The novelty lies on the improvement of the numerical model, to make it suitable for early warning and to have determined the best set of parameters so that they don’t need to be calculated when an earthquake occurs. Also, it proposes a workflow considering the times available for risk disaster management.
Another concern deals with the poor literature review. Only 18 papers are cited. The introduction does not present the state of the art of tsunami warning systems. With a more comprehensive literature review, the reader could better understand the importance of the proposed methodology. Other tsunami warning systems may be described and not only the Chilean case, such as the discussion can clearly state the advantage and disadvantage of the proposed methodology (What about the warning systems in Italy, Indonesia, US, Japan, etc). Despite the study is focused on the Chilean case, the tentative findings could perfectly be applied to other regions. In addition, important papers are missed such as Cienfuegos et al (2019), Selva et al (2021), Behrens et al (2021).
Introduction has been improved, cinluding experiences from other Tsunami Warning Centers.
Another concern deals with the model validation. Lines 228 mentioned that finite fault solutions of 2010, 2011, 2014, 2015 and 2016 earthquakes are the input in the numerical solver. However, it is not mention which finite fault model were used. Are they your own proposed finite fault models? If so, they should be based on the preliminary seismic information to demonstrate that your early tsunami warning system is successful. Or are you just using other authors model to validate the numerical model?. If so, I am confused about the focus of your paper. Are you just proposing a numerical model or are you improving the warning system?
First the Finite Fault models used are taken from the National Earthquake Information Center at the United States Geological Survey (clearly mentioned in Line 240). We are not validating the seismic source, we are using the fastest seismic source available, but to validate the model a final source solution was used. The comparison between earthquake sources, is out the scope of this paper.
The quality of the paper and written English is very poor. It is important editing and proofreading before manuscript submission.
Ok, improved.
Section 2 “Methodology” described only the numerical model for tsunami numerical simulation. What about the finite fault model solution? It should be part of the methodology as well.
No, it’s out of the scope of the paper. This was improved in the manuscript to make it clear.
In section 3 related to the validation of the model, I strongly recommend using the tsunami record at Coquimbo tide gauge. Why is it avoided?
It is included now
Line 216 mentioned an elliptical finite fault model, but there is no description of this model in the methodology. How is this model generated?
Now it is cited, Fuentes 2019.
Line 218 mentioned that “after few minutes the FFM is expected to be computed…” but Figure 4 shows the FFM from USGS. I would expect you to show your own finite fault model computed with available information just few minutes after the earthquake.
We are not calculating finite fault. We are showing the advantages of this particular tsunami modeling. Clarified in text to avoid confusions.
Line 229, which finite fault models were used? The 2010 has 19 proposed models. It is necessary to indicate the authors in each model. A table would be acceptable. In addition, an explanation on why each model was selected is necessary.
First there are more than 19 finite fault models. Here we use the USGS FFM official solutions, as stated in different parts of the manusctipt.
Figure 5. It is necessary to add Coquimbo and Valparaiso to the validation of the Illpapel event. Why are these locations discarded?
Valparaiso does not have data for the first two hours. Coquimbo is now added
Line 247, “…different parameters were used,…” Which parameter are you talking about.
Simulation time, spatial resolution and grid extension. Clarified in the text.
Figure 7 shows different tsunami hazard levels, but theses levels are not defined previously. Is it related to Table 1? Then, it should be mentioned in a Methodology section.
They are defined in table 1. Now its referenced before figure 7.
Line 249, “…while figure 7(f) shows the tsunami threat evaluation for far field…” is it Figure 7(g)?
Right. Corrected.
Line 272, “several points were created…” where are these points located? At a given water depth of at the coastline?
On the coastline, more precisely on coastal boundary condition points on the grid. Clarified on the text.
Figure 8 is really confusing. Which is a, b or c?, Why the description mentioned twice c? one is with 5h simulation and the other is with 12 h simulation.
(a), (b) and (c) added to the corresponding figure. The description was improved.
Line 296297, “…the seismic source information…” what is the uncertainty related to those parameters? There is no explanation in the paper related to the finite fault model once the seismic parameters are computed.
Its out of the scope of this paper. We are showing the performance of the tsunami modelling, not assessing the seismic source methodology.
Line 332, “we model the fault plane defined by the Wphase moment tensor”, but this is not explained in a methodology section. It is necessary to describe the methodology and the time needed to be computed. What is the uncertainty related to this preliminary fault model?
The methodology corresponds to Fuentes et al (2019). Now its mentioned. The uncertainty of the seismic source is out of the scope of this paper.
Line 333, “Once the finite fault model is available…” what is the uncertainty of this model, and how long does it take to compute this fault model?
The uncertainty of the FFM is out of the scope of this paper. The computation time is mentioned in figure 9.
Line 340, “…uses a point source as input”. It should be explained in a methodology section and the time needed to be computed must be mentioned as well.
The computation time of the point source solution is less than 5 minutes, as mentioned in figure 9. The methodology is out of the scope of this paper.
Figure 9 should be in the results or a methodology section. If this figure is proposed in Riquelme et al (2021), I do not really understand what the novelty of this paper is.
Figure 9 is not proposed by Riquelme et al (2021), he only mention the computation times of each seismic source, which has nothing to do with the tsunami model presented here. Figure 9 is the proposed workflow, using each seismic source solution from the Chilean Seismological Center,
Citation: https://doi.org/10.5194/egusphere20233061AC1

AC1: 'Reply on RC2', Matias Sifon, 29 May 2024

AC2: 'Reply on RC1', Matias Sifon, 29 May 2024
The This is a meaningful study, but its significance lies mainly in the practical tsunami warning operation rather than the science. As a reviewer, I believe that the research has limited scientific innovation as it does not propose its own theories or methods. The methods used in the paper, whether Okada's model or the nonhomogeneous linear shallow water equation, are already mature research outcomes. The primary work of the authors involves discretizing these equations, writing them into programs, and applying them to practical operations.
This work is important, but similar models have been proposed by operational departments in various countries. The authors should at least elucidate the advantages of their method compared to others, such as the one adopted by PMEL, in order to distinguish their approach.
At the same time, I am also curious whether this method has the capability for adjustments during its operation. As shown in Figure 5, the method exhibits insufficient fitting for some stations at the first peak. Can existing models be corrected through observational data?.
It can be corrected as soon as more data comes in. However the scope of the paper is to enhance the tsunami early warning system. We trade off accuracy for speed. Then even this is possible, the idea is to model the tsunami before the first arrival. So, if the data allows the seismological agency to model the earthquake source before that time, it can be corrected, if not, we get the model with the information available at that time.
Overall, I believe that this manuscript is not suitable for publication in NHESS. Unfortunately, I would like to reject this manuscript. Please consider submitting to a more suitable journal. The following comments are for the authors' reference, hoping they can further refine this study.
Line 25: Please change “y” to “and”. As a manuscript submitted to an international journal, it is important to use English expression.
This is a typo, done.
Line 25: The reviewer believes that using the term “sensitive” is not appropriate. Chile is relatively close to the subduction zone, but other countries (such as Japan and Peru) also have similar situations, which cannot be called “sensitive”. Additionally, some areas may face tsunamis triggered by landslides, with similarly short reaction times.
Thanks for the suggestion, done. (highly vulnerable to tsunami threats)
Line 35: Why does the system only consider subduction type earthquakes? Are there any technical issues in considering other types of earthquakes?
There are no limitations. This can be done for any type of earthquake, but since we focus on Chile, and there are not normal or strike slip events that have generated tsunamis records on sea level instrumentation, we used the ones that have data, which are subduction type events.
Line 50: Although one can understand the meaning of each physical quantity after reading the formulas, I suggest the author explain the meaning of each physical quantity to enhance the readability of the article.
We have explain each term of the equation in the revised manuscript
Line 70: The assumption of null time derivative/velocity is based on what evidence? Please specify.
In the standard modeling. The principle of causality nothing happens before that. Typical modeling.
Line 96: What is the scale of the damping layer? How many grid points does it occupy, and has sensitivity testing been conducted?
We just added a paragraph discussing this issue.
Line 138: Why is Coriolis force not considered? I understand that the author is mainly calculating nearfield tsunami, but if considering propagation over slightly longer distances, such as Easter Island, Coriolis force needs to be taken into account.
We are focused in the near field tsunami modeling. Eastern Island or such locations can be modeled properly with all the terms because this is not near field and the arrival time is not a requirement. The tsunami arrival time could be several hours. A clarification was added.
Figure 5: The figure title mentions four events, but please indicate them in the figure.
This was included in the revised manuscript. The figured has colors for each event.
Citation: https://doi.org/10.5194/egusphere20233061AC2
Viewed
HTML  XML  Total  BibTeX  EndNote  

235  88  22  345  10  16 
 HTML: 235
 PDF: 88
 XML: 22
 Total: 345
 BibTeX: 10
 EndNote: 16
Viewed (geographical distribution)
Country  #  Views  % 

Total:  0 
HTML:  0 
PDF:  0 
XML:  0 
 1