The manuscript addresses a highly relevant topic for the natural hazards and coastal risk community. It makes use of data-driven metamodels as an effort to support operational early warnings for marine flooding. The comparison of analog (M1) and meta-model approaches (M2, M3) is scientifically interesting and potentially valuable for researchers and coastal managers. Furthermore, the study is generally well structured and the outcomes are sufficiently described.

However, in its current form, the manuscript requires further clarifications and editing improvements before considering publication. My main concerns arise from the following aspects:

(1) The authors refer to “overflowing” processes, but the widely accepted term of wave run-up is not introduced or discussed. This aspect should be clarified and integrated into the terminology of the manuscript.

(2) Several aspects of the modeling strategy require clearer justifications, including a brief presentation of the constructed pseudo-historical events. In particular, the construction of pseudo-historical events should be better described, including the ranges of forcing parameters. Furthermore,

(3) The extraction of the surge level (SPM ) parameter took place at the location of the buoy located offshore (at 50 m deoth) and not at the tide gauge located at the nearshore.

(4) M3 method should be more clearly framed as a refinement or extension of M2, mainly improving the spatial reconstruction of flood depths rather than the prediction of flood extent or impact category.

(5) The inclusion of 1000-year return period storm scenarios raise concerns, given the the limited observational basis and uncertainties of the hydrodynamic parameters, and thus should be reconsidered.

(6) Use of terminology, technical specifications and text editing.

Once these issues are adequately addressed, I believe that the study could be a meaningful contribution to the coastal hazards literature. For this purpose, detailed commentary is provided to the authors (see supplement).