Modelling with Volna-OP2—Towards Tsunami Threat Reduction for the Irish Coastline

Round 1

Reviewer 1 Report

Generally, the research looks at the sue of Volna-OP2 to model a tsunami threat to the Irish coastline. The manuscript itself was an easy and relaxing read compared to other research papers on modelling tsunamis. However, there are a few major comments and suggestion that the authors must address:

• The introduction is lacking in terms of discussing previous major tsunami events such as the 2004 Indian Ocean Tsunami and the 2011 Great Northeast Japan Tsunami and the lessons learned from these. This includes the changes in predicting and modelling tsunami threats. What changed? What is lacking from these changes? How did you address it? 
• The introduction section needs to discuss how your study compares to other previous tsunami modelling efforts, both within and outside your region. This may help the authors highlight the significance of this effort beyond stating that this is a pilot study.
• Provide a map of Ireland, highlighting where you key points of interest are. This is to help the study easier to read for those who are not familiar with Ireland's geography.
• Section 3.1 is a bit off in the structure of your study. While it is used to present Volna-OP2's ability to predict rapid regional forecasts, there is not much discussion more than that.
• Related with the previous comment, it is suggested that the authors change and modify section 5 into a discussion section, tying the previous sections together. Moreover, following your results, relate back to previous studies. How does it take off from previous work? How does it differ? How can disaster management use this?
• Conclude by briefly recapping the merits of Volna-OP2 and its use in tsunami forecasting as based on your study and its implications to the field of research in general. 

In itself, the review may look a bit daunting since it calls for an added section. However, the discussion and merit of the study is already within the text. It basically needs to provide its significance to the larger tsunami research field in general.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The Geosciences manuscript with title “Modelling with Volna-OP2: Towards tsunami threat reduction” by Giles et al. presents results using the parallelized version of tsunami model VOLNA named VOLNA-OP2. The authors showcase the capabilities of the model using two case studies: one historical tsunami for the Mediterranean Sea and one for the North East Atlantic. The Boumerdes 2003 historical tsunami was chosen for the Mediterranean and the authors showed how the model can be used to obtain faster-than-real-time simulation results to account for the uncertainty of the seismic source for tsunami early warning operational forecasts. The historical Lisbon 1755 earthquake and tsunami was used as a scenario for the North East Atlantic. The authors chose six earthquake source parameter sets for this event from the literature and ran wave propagation simulations for the North East Atlantic region to identify the source that has the biggest impact on the Irish coastline. High resolution unstructured meshes of ~10m resolution near coastline were created for two study areas, Galway Bay and Dunmore East, to produce tsunami inundation maps using the scenario producing the largest wave amplitudes near the study areas. After running the code using the optimum CFL number, the inundation zone and maximum wave runup were computed for each study area to draw conclusions on the vulnerability of these areas to earthquake-generated tsunamis, faults offshore Portugal that produced the 1755 Lisbon tsunami being considered the main source of tsunamigenic earthquakes in North East Atlantic. 

While the usefulness of such a faster-than-real-time numerical model for tsunami early warning operational forecasts is not in doubt, the manuscript does not offer any new or useful information. The chosen application examples in my opinion are neither complete nor well presented. In the case of the Mediterranean simulations ensemble, the authors do not present a methodology to deal with source uncertainty, but they simply present max wave height plots for 20 source parameter sets unknown to the reader. The important information gathered from this example is the runtime for the complete set of simulations, with the runtime not being compared to the serial code for context. Similar results are presented for the North East Atlantic scenario, with the parameters of the six selected sources outlined in a table. For the high resolution inundation simulations in the two study areas in Ireland, results were presented only for the scenario producing the highest amplitudes, which nullifies the purpose of using a fast-running code; fast codes are typically applied to run many cases to obtain the uncertainty space. The tsunami inundation extent and maximum runup in the study areas were obtained for the the 1755 source producing the highest amplitudes. However, such information can be obtained from any benchmarked tsunami code and therefore should not be the main outcome of a publication presenting the capabilities of a new faster-than-real-time code. In my opinion, the authors should focus on one application that really showcases the capabilities of the model and present it thoroughly, highlighting all the information that would make such an application difficult using a serial code.

For the above reasons I’m afraid that I have to recommend this manuscript not to be accepted for publication in Geosciences and be resubmitted after making major changes. See more comments below.

________________________________________

Major comments

- For the Boumerdes 2003 case, what is the added value and which capability of the model is illustrated through this event that is not shown with the Lisbon 1755 event? The way it is presented, the only difference is that 20 simulations were ran instead of 6. But there is much more to it when it comes to tsunami early warning applications. 

- In the high resolution inundation mapping for the 1755 Lisbon tsunami event, the authors only present results for a single source, which is not an ideal way of presenting the capabilities of a fast code. This is a typical application of a serial code which would yield similar results if setup correctly. The runtime information was not compared to a serial code which would highlight the benefit of using such a high performance code.

General minor comments

- Figure quality can be improved - see minor comments below.

- Writing quality is good but many commas missing in sentences. 

Other minor comments

- Line 19: I don’t see what the rarity of tsunamis has to do with the tools used for accurate forecasts.

- Line 30: The limitation is not the modeling software per se, but the computational time needed to run the modeling software. It should be better worded. 

I would also rephrase the part “…for both TWCs…” as “…for both TWC operations…” or “…for both real time tsunami forecasts…”.

- Line 39-41: After reading citations #7 (Reguly et al., 2018) and #8 (Giles et al., 2020), citation #7 seems to be a more appropriate reference for the benchmarking of the updated code.

- Line 83: add comma after “As stated”.

- Line 95: I would suggest editing the section title to reflect that the Boumerdes 2003 tsunami is simulated - same applies to the North East Atlantic Tsunami.

- Line 96: add comma after “of example”.

- Lines 101-102: only earthquake magnitude was varied or other source parameters as well? 20 different initial earthquake sources were identified in the literature? Not clear from the text.

- Lines 105-107: It is not clear to me what the 2arc minute grid corresponds to. Is that the resolution of the input bathy/topo grid or the actual computational grid? I thought the code is using unstructured grids. Please explain in the text - same applies for the North East Atlantic Tsunami case. 

Edit: line 66 says the code runs on unstructured triangular meshes and the reader has to get to line 154 to realize that these simulations were ran on a uniform grid.

- Line 111: how can the reader see that if the earthquake magnitudes corresponding to each subplot are not stated in figure 1?

Although in most cases it is true that larger magnitudes lead to larger tsunami amplitudes, it is not true for all cases. Different source depths, slip magnitude/rupture area ratios result in different max wave height distributions. This sentence needs to be rephrased.

- Lines 113-114: the mean is the most probable distribution of maximum wave heights only if all scenarios are assigned equal weights/probabilities.

Also, the absolute maximum wave heights resulting from the 20 simulations are the worst case scenario of the scenarios considered. Can’t make the claim that it is the worst case scenario in general. For the North East Atlantic Tsunami the words “rough estimate” were used to reflect that.

-  Lines 121-123: add reference(s) for the Lisbon event.

- Line 127: add comma after “in mind”.

- Line 128: which sources were used? Provide references.

- Line 133: delete “local”.

- Line 141: add comma after “perspective” and replace “Lisbon type” with “Lisbon-type”.

- Figure 3: the authors should consider plotting the figure using a Mercator projection (e.g. using mapping software such as GMT), otherwise the Irish coastline appears deformed (high latitudes). 

- Line 158, Table: add the earthquake magnitude for each source and state in the table caption that these sources correspond to the Lisbon 1755 earthquake.

- Lines 181-183: it is not clear what the authors are trying to say here. That they did not have topo data to present results for other coastal areas ?

- Figure 5: big fonts used for x-axis and y-axis legends, and very small fonts for the tick labels, to the point one can’t read them. 

Top subplot titles refer to M5 and M3 without any context. Which wave gauge is M3 and which is M5?

The rest of the subplot titles refer to locations that the reader has to search online to find where they are. While I understand that each of the subplots is intended to be used for relative comparisons between time-series produced by each source, the absolute amplitude values are meaningless unless the location and local depth are stated/shown.

The authors should label the gauges in Fig. 3. Text to show which rectangle corresponds to Galway Bay and which to Dunmore East can also be added.

Last, Fig. 3 shows numerous (I’m counting at least 13) wave gauge locations, and only 6 are shown in Fig. 5 - same for Fig. 6. Very confusing. 

- Figure 6: wave gauge locations are hardy visible and font appears skewed. 

It would be more useful to use the title of each subplot to label the source number instead of repeating “Maximum wave height and areas of interest”.  This information can be seen from the colorbar label and figure caption.

- Line 188: the authors have to elaborate more for this information on the datum and coordinate system used for the topo/bathy data to become useful to the reader.

In general, much more work than just a datum adjustment is needed to get a seamless relief grid from combining topo and bathy data. 

- Line 190: text doesn’t refer to Fig. 8. 

- Line 191: typo for word ‘splits’

- Figure 8: The gray outline of the mesh does not allow to visualize the characteristic element size in the high-res areas. It just appears gray.

- Line 202: delete “for” before “this”. 

- Line 206: add comma after “Further”

- Line 210: location of River Barrow should be shown in Fig. 10.

- Line 241: add comma after “To date”.

- Line 270: hyperlink not working.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors properly responded to the reviewer’s comments and extensively revised their manuscript. 

The new manuscript is much improved and the scope of the paper is much more clear now.

I have included some minor comments below for the authors to consider. 

________________________________________

General minor comments

- Figure quality can still be improved. Figure font and symbol sizes should be checked to see if someone can read them. For example, in Figure 2 of the revised manuscript, the symbols and text in the map need to be much larger for anyone to read them off the printed figure in an A4 paper. For Figure 5: I don’t see any wave gauges highlighted in any of the subplots. Also, the last sentence in the label of Figure 5 can be removed now that the titles of the subplots have been replaced. Check all other figures for similar “weaknesses”.  

Other minor comments

Point 2 – In the high resolution inundation mapping for the 1755 Lisbon tsunami event, the authors only present results for a single source, which is not an ideal way of presenting the capabilities of a fast code. This is a typical application of a serial code which would yield similar results if setup correctly. The runtime information was not compared to a serial code which would highlight the benefit of using such a high performance code.

Reply – We now present high resolution inundation results for all of the four sources investigated. In order to achieve this in the short time-frame, we have carried out ’hotstart’ simulations, where the output from the coarser mesh is interpolated onto the fine mesh and acts as the initial condition. A full explanation of this procedure is outlined on lines 209-211. We hope that the reviewer finds this suitable. With regards to the runtime comparison to serial codes, we have carried out and included the relevant runtimes for the serial and MPI/OpenMP version of the Volna-OP2 code where applicable. Details on these runtimes can be found on lines 135-136 and 173–175.

Reviewer reply - I do find it satisfactory, although I don’ t understand why the authors did not run all four simulations using the fine-resolution grid from the start of the simulation. If 6 hours of simulation take 2 hrs computational time, 10 hours of simulation would take a mere ~1.33 hrs more computational time. Did the authors have operational applications in mind when deciding to go that route?

- Lines 197-204: The authors provide many more details for the high-res grid development in the revised manuscript. However, the issue of tidal level has not been raised in the manuscript. For example, a quick search reveals that the tidal range in Galway is of the order of 5 m. Therefore, the tidal level considered in the simulations would have a big effect on the simulated tsunami inundation extent. There needs to be a sentence or two after line 204 discussing this. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf