| 09 Sep 2026
Historical Marmara Tsunamis Reconstruction Through Physics-Based Scenario Selection
Abstract. The Marmara's Sea with its near-field tsunami complications in Europe and the eastern Mediterranean, is paradoxically surrounded by a densely urbanised, infrastructure-rich coastline and positioned above the North Anatolian Fault system with historical earthquake events, submarine slope processes and narrow-basin resonance that impact coastal water levels. This fact motivated the current study presenting a benchmark-consistent GeoClaw modelling roadmap on seven historical Marmara tsunami scenarios selected from 22-event source catalogue with numerical comparison through 20-lane. The lanes represent exclusively, different source-construction, turning strategy, comprising earthquake-only, earthquake-plus-landslide, adaptive sampling, hybrid site-blend, and close-loop correction variants. From the first lane to the final selected best lane of the simulations set, the benchmark error could be reduced from maximum 1.93 m to 0.019 m in MAE (Mean Absolute Error) and 2.26 m to 0.024 m in RMSE (Root Mean Square Error), with all seven events cases remaining below 10 % error and the largest absolute percent error is 2.4 %. After target-based scenario selection, the corresponding outputs such as gauge-history, travel-time, and bridge-ready hydrodynamic outputs, providing a reproducible chain from historical source interpretation to infrastructure demand inputs are extracted. The workflow is designed for regions where instrumental tsunami records are sparse, but historical runup constraints, bathymetry, mapped faults, and scenario logic can be combined transparently.
This manuscript presents a GeoClaw-based method for reconstructing historical tsunami events in the Sea of Marmara. It narrows 22 possible reconstructions to seven by comparing modeled runup with historical runup estimates. I think this approach could be useful for comparing possible source mechanisms, but I have strong concerns about how the method is currently described and applied.
My main concern is whether the proposed sources are physically plausible. The reconstructed earthquakes are strike-slip events with magnitudes below Mw 7.5, which seem unlikely to generate the modeled tsunamis on their own. The methods refer to “NHESS-backed” reconstructions, but do not explain how those sources were constructed or whether they include other mechanisms, such as landslides or significant seafloor uplift. The focal mechanisms shown also do not appear to include substantial normal or thrust components. The “20-lane” analysis seems to move away from earthquake-only sources, but the discussion identifies one selected source as an earthquake-plus-landslide scenario and describes the others as “NHESS-backed” without defining that term. I would like to see the source construction and selection process explained more clearly.
I also have concerns about the EQ1 results. The source fault is described as being on land west of the Sea of Marmara, but the initial sea-surface perturbations appear on the far eastern side of the sea. Since EQ1 is used throughout the results and discussion, the authors should check that the correct data are plotted and explain this mismatch.
The methods need more detail so readers can understand and reproduce the process. Some figures are also difficult to interpret because the captions do not explain enough and the labels and legends are too small. There are additional grammar, tense, and word-choice issues that should be cleaned up. Those edits are secondary to my concerns about the sources and results, which affect my confidence in the method and the paper’s conclusions.
The clarification that these outputs are scenario products rather than unique historical reconstructions appropriately narrows the manuscript’s claims. However, it does not address whether the source scenarios are physically plausible. A low runup error alone may favor a source that fits the estimates but is not supported by the tectonic or landslide evidence. Physical plausibility should be an explicit constraint on scenario selection, or the manuscript should clearly distinguish physically supported scenarios from those included only because they fit the runup estimates.
I've attached my annotated pdf file where I placed comments.