The paper by Kiesel et al. presents a modelling framework to assess the extreme sea levels and the extent of coastal flooding on the German Baltic Sea coast. The topic of the paper is highly relevant and is of great importance for the planning of the protection of the coastal regions. On the Baltic Sea coast, high sea level extremes are possible and will become more frequent with the global mean sea level rise. The paper is well written, the methods and results are clearly described. The quality of the presentation is high and the text is understandably written using good English.

I have only some minor comments for the authors to help them clarify some details.

Line 84 Explain what a hydrograph is, term might not be familiar to all readers

Line 167 Is the boundary condition the same for neighbouring boundary points? I understood from the text that the boundary point has as its boundary condition the hydrograph from the nearest boundary station. This could be explained in detail in the text.

Line 234 Has the land uplift due to postglacial rebound taken into account when SLR has been subtracted with a linear fit? You could mention the land uplift rate on the German Baltic coast and discuss whether it is relevant in this study.

Table 4 and Figure 5

It would be nice to have the station names of Table 4 in Figure 5 to be able to locate the stations in the map. The station could be given a number which is shown in Figure 5 to avoid too much text in the Figure. Figure 5 could also be larger, because it is one of the most interesting figures in the paper.

Line 357 You could discuss why the peak water levels are higher in SH than in MP in the 200-year event. Is it due to the shape of the coastline, does the bathymetry affect it?

The manuscript describes a modelling framework for the German Baltic coast to assess coastal extreme sea-levels and associated flooding. Such a tool is of great utility for coastal management and planning in view of sea-level rise. The methods employed are within the state of the art (numerical modelling tools, analysis methods for EVA, etc) . The framework is not new in terms of these methods (multiple similar systems have been developed in other parts of the world) but it is new and the first of it's kind in the specific region of the German Baltic coast, according to the authors. As such, it provides the opportunity to bring new knowledge on coastal flood risk in this region. However, I think the manuscript should be re-structured to better highlight its objetive. It appears that the manuscripts aims to describe and validate the modelling framework but also demostrate its use for a specific application, which seems to be coastal protection design (given the recurrent link to the 200 year RP design level and the simulations performed for this RP). Or it could be that the objective is mainly the design level calculations and associated flood charachteristics, and the validation is just a necessary step? This is just my interpretation, but whether it is correct or not, the message is that the objective is not clearly presented in the introduction and followed through in the rest of the document. For example, a proper carachterization of flood charachteristics - which is what the title of the manuscript implies -would require the assessment of multiple return periods, while here results focus (only) on the 200 RP design level. I therefore recommend to restructure the manuscript, and rename it if necessary, according to the objetive and research question at hand. This objetive should be clearly stated in the introduction, before elaborating on the methods used and simulations performed (which should be justified by the chosen objective), and it should streamline the rest of the document. Because of the need to restructure the contents, I propose a major revision. That said, the manuscript is scientifically sound, so no new results should be needed to get it to a publishable level.

Additionaly, harmonization is needed relative to the use of the terms waterlevel, surge etc. This relates to some specific comments pointed out below. As it is, it is not clear in the document if the return periods derived correspond to waterlevels (including tides) or just surges. The term waterlevel is mostly employed, but according to the text the hydrograph definition is based on surge. This aspect should be clarified in the methods.

In the following, I elaborate on some more specific edits needed in each section:

1. Introduction

A list of difficulties faced during "large" scale coastal flooding assessments is presented (resolution and computational burden, complexity and sensitivity of models, lack of accurate data on flood protection features like dunes and dikes, validation material, uncertainty in extrapolation to high RPs, etc), and it is followed by 'Here we address these challenges by developing a new modelling framework' (line 74). Again, this is not the objetive of the paper, as it doesn't provide a solution to lack of computational resouces to reach relevant flood scales at regional scales (for example showing a faster modelling framework) , sparseness of data, etc. I think this part of the introduction should be rephrased to highlight that these typical difficulties, which hamper the validity of broad scale flood assessments, have been taken into account in the design of the current modelling framework, by collecting data on hard structures, using a high-resolution DEM, etc, which were fortunately accessible to the authors for this specific region. These ascpects remain a challenge elsewhere. In terms of validation, the usability of SAR is explored as a promising (and emerging) data source.

If the manuscript is reframed to target the dike/coastal protection design application (even if as a showcase for the framework), I propose that the associated, area-specific information presented in section 2 (lines 107-113) is moved to the introduction as part of the contextualization of the question and objetive at hand. 

1. Methods 

Line 124: Please include a proper reference to the IPCC report mentioned. 

Line 132: correct hind-cast to hindcast

General comment in section 3.2.1: Information on the modelled physical processes is missing, does the model include tides (at the boundary)? What atmospheric parameters are used for forcing (wind is mentioned, but what about atmospheric pressure?) Also, 2 different forcing products are used for hindcast and for the 2019 event, what are the spatial/temporal resolutions of these products? (as this has a mayor impact on storm-surge extremes). Some calibration of bottom friction is mentioned in order to increase extreme surge heights, and hindcast winds have been increased also to increase the surge. Little information on these calibration exercises is provided, could you elaborate a bit more? It is important to refer to this in the discussion,as it is mentioned that waves setup was not included because the model showed already an overprediction of extremes....

Section 3.3 -->Rename to 'Sensitivity analysis'. A calibration would entail benchmarking against the truth (e.g. observations) to choose the best settings, but this is not done.

Section 3.4 --> There is too much detail on the GEV method (e.g. thedescription of subfamilies), a proper literature reference should suffice (as it is a very widely used method). The generation of the hydrographs constitutes a more novel component here, consider adding in this section an example for the generation of the hydrograph and the scaling to a given RP. In relation to the generation of the hydrographs (Appendix A), please include a justification for the choice of 1meter as threshold to isolate extreme events. In typical peak-over-threshold (PoT) extreme value sampling, a high percentile is often used, which ensures a minimum sample size. Do you use meteorological independence criteria between events (as typically done in PoT)? Is the 6 day event duration (3 days before, 3 days after) justified somehow for this region?

Line 225: replace 'stretched' with 'scaled'

Line 226: replace 'than' with 'then'

Section 3.5 --> It would be good to include a spatial map of the TGs used, with a number asigned to each, such that these identifiers can be used throughout the body of the text (for example when mentioning a TG, with name[number] ) and in subsequent plots. It is difficult to follow otherwise for someone not familiar with the region. For the EVA, it would be interesting to see for those TGs where your records are longer than in your hindcast, what is the impact of clipping the TG series to the hindcast period on the EVA? You could be sampling more (and larger/smaller) extreme events in your TG than in your model...which might (partly) explain differences in the fits.

1. Results and discussion

-Jusitfy the choice of 1meter as threshold for extremes in the hindcast validation (as for the event isolation in Appendix A, see comment before)

-Near coastal lagoons, the model has been deemed inapropriate and TGs have been used instead for the flood modelling. This might be justified for this particular application or assessment, as an ad-hoc solution, but it is not part of the framework per se. This should be discussed in section 4.6. 

-In line 210, 30 tide-gauges are listed, but only 28 appear in Figure 6. Why?

-The empirical values in the distribution shown for a TG (Figure 4) show an increasing bias for increasing RPs, but the fit seems insensitive to this. Could you elaborate? This seems also evident from the 2019 event, what is the corresponding RP and how does the error compare to the derived RP error, given that a different meteo forcing is used? The mean (negative) bias for the 2019 event is partly compensated by the positive bias in the stations around some of the unresolved lagoons, you could leave these out to have a better picture of the model performancefor this event? (since TGs are used for the flooding at these locations anyways)

-For the Hit/False/Miss score assessment, could you not interpolate the 50m modelled fields to the SAR resolution of 10 meters? The score percentages are confusing as they are

Lines 355-359: There is a lot of speculation in these lines. Multiple instances of 'We believe that...'. Avoid speculating, you have all the results at hand to do an analysis of what is leading to differences between regions.

Line 370: I find this discussion bizarre, and the purpose of analyzing the linear correlation with peak WL is not explained. Why would we have a linear increase in flooded area? That depends on the inland topography, the hydraulic connectivity etc, no? why are you showing this?

Line 382:To the knowledge of the authors, this study constitutes the first regional-scale assessment using a high resolution, fully validated and offline-coupled modelling framework that incorporates natural and anthropogenic flood barriers to assess extreme sea levels and associated coastal flooding along the German Baltic Sea coast. This should be in the introduction, together with a clear objective.

Line 388: Gulf of Finland and Florida-->It is a strange mix of locations, I suggest to remove Florida and focus on Finland which is at least in the Baltic.

Line 390: ‘in the previous location’ rephrase it is not clear which you refer to.

Line 394: : first, wave setup is included in 395 the tide gauge records that we use to extrapolate the 200-year return water levels. This is a bold statement, up until recently it was argued that TGs didn’t record wave setup because of their sheltered location in harbors, but recent studies have shown that sometimes this assumption doesn’t hold. Do you have prove or can you reference a paper where this is evaluated for this region? Otherwise rephrase.

Line 395:  for 21 of the 30 tide gauges, the coastal ocean model still overestimates the 200-year return water level. Elaborate on why this is the case, while for the 2019 we have a general underestimating trend, and it has been shown that extremes are generally underestimated in Fig 3. It looks like the EVA plays a role here? The model was calibrated also to increase surges (increased wind, reduced bottom friction), still it seems to generally underestimate, but after the GEV fit it overestimates? Using this same argument, and if one only looked at the hindcasted extremes (actual modelled extremes), one could advocate to add wave-setup instead in order to increase the underestimated surges...

Line 396: . Second, there is still no conclusive information on potential changes in wave climate, and results show strong spatio-temporal variability (Weisse et al., 2021). I don’t see why this is a reason not to include wave setup in your assessment, which is based on a hindcast and not projections. You are also not evaluating (future) changes in storm surge, but you still include this process.  

Line 397: Finally, potential wave setup along the German Baltic Sea is arguably low compared to uncertainties associated with the simulated SLR projections. Do you have a reference for this?

Line 417: associated do you mean ‘future’

Some limitations haven’t been discussed: Lack of processes in the hydrodynamic model (e.g. mean sea level variations), lack of resolution to solve some regions, source of general underestimation of extremes, the uncertainties in the EVA method (especially when looking at such high return periods, general rule of thumb is that records can be extrapolated via EV distributions up to 3 times their length). The hydrograph method must also introduce a lot of uncertainties, and I presume the flood maps are quite sensitive to the choice of threshold used for the hydrograph design as well as to the duration of the event, now fixed at 6 days. How does this compare to the duration of the event during the January 2019 storm? How does the hydrograph vs time-series approach compare for this particular event, when you use the corresponding return period?

1. Conclusions

Modify according to the objective of the manuscript. If the objective is to present the framework, include other possible applications, and future directions of development based on the current assessment limitations and envisioned applications. If the objective was the 200RP flood characteristics dataset, explain how it can be used for spatial planning, for (climate) dike design, etc. Some of these aspects are now briefly mentioned between lines 430 and 435 but without a clear objective stated for the paper at the beginning, it is difficult to understand.

The manuscrit by Kiesel and co authors adresses the difficult problem of coastal flooding at regional scale along Germany Baltic Sea with multiple difficulties, i.e. data resolution, account for dykes, validation, cascade of uncertainties, etc. The authors propose a modelling framework to do do with a key ingredient being the use of mention images for validation of the assessment.

I believe that the results have the potential to be of interest  for a wide audience. Yet several aspects remain unclear and I recommend major corrections before publication.

1) Multiple problems are tackled but with different degrees of achievement, and as reader we get lost about the main message. Is the main result the framework ? If so, a figure presenting the different steps would help a lot together with a discussion section dedicated to the limitations of each this step. If the main result is the use of satellite images, this should be reflected in the title as well in the introduction.

2) the pre-treatment of the data for extreme value analysis is unclear to me. The authors use the notation 'esl' which gives the impression that the authors work with total water level. However the authors also mention extreme surges. If this is the second option, how do the authors derive the total water level at the coast to force the flooding model? More speciffically how do the authors handle the convolution with tide?  In addition do the authors use skew surge or the instantaneous surge?

3) the analysis of fig. 4 shows some similarities of the enveloppe of uncertainties. Could the authors provide more details about its computation. Finally what surprises me is that there is quasi systematic underestimation of the empirical points except for the more extreme points which is overestimated. Not having this last point would change completely the analysis. Could the authors comment on that?

4) The analysis of the differences with the image data (fig. 7 in particular) is of high interest. Could the authors elaborate more on the added value of their updated dem with dykes information. Would it make sense to also compare the results with the ones of a traditional approach without this information? In addition, would it be possible to plot some examples of dem in the vicinity of the dyke to picture how much correction should be done?

5) the approach described in appendix A is interesting to account for the time evolution around the surge peak. Given the variability shown in panel a) of fig. A1, could the authors comment on the possible impact of using only the mean time signal instead of a model accounting for this variability?