I commend the authors on this mammoth work on a topic that is highly relevant and much needed for Bangladesh and this region. The work has been performed well and described thoroughly. I believe that the paper will be ready to publish after the authors clarify a few doubts and questions.

The issues that require clarification center on two points: 1) the embankments ; 2) the cyclone ensemble and return periods.

Re. the embankments:

1. The authors should explain exactly how the information on embankments has been integrated and should provide a map and/or more information showing embankment heights within their dataset (this can be smoothed out if there are copyright or other data concerns). As the authors state, these embankments are critical in controlling flooding. However I do not know if these embankments are wider than 250 m, and it is not clear if/how these can be captured by the bathy-topo datasets if the minimum resolution is 250 m.

2. The authors mention that the embankments appear to start overflowing at the 75-100 year RP water levels. Clarification on how the model "sees" these embankments will be useful. I assume that overtopping processes are not included in the model but this will be worth stating.

Re. the cyclone ensemble:

3. I understand and support the authors' decision to show water level variations based on RPs rather than events. While logical, this can however be confusing to interpret. The authors should add a few sentences explaining describing how the 100 year RP water level map (for example) is comprised of WLs from several cyclone events.

4. Figure 4 indicates a significant spatial variation in the number of cyclones affecting the coastline, with the western coastal segments seeing a lot more cyclones. This could mean that coastal segments in the east, with far fewer cyclones, also have a much smaller range of WLs within the 5-500 year RPs. Can the authors clarify if this is the case, and whether/how the spatial variation in cyclone tracks can influence the spread and uncertainty around the WL results at the shoreline and inland?

The manuscript by Khan and co-authors addresses the problem of storm surge hazard assessment over Bengal delta. They propose a hybrid procedure that combines high resolution numerical simulations, ensemble modelling, and probabilistic analysis. Their results are of high interest for coastal risk planning in this region both for present-day and future climate conditions.

Main comment

The manuscript is well organized and the presentation of the methods and results are very clear. The conclusions are sound. I very much appreciate the efforts made by the authors to discuss their results with respect to existing studies and to the limitations of their work (Sect. 5). My background is mainly focused on statistics. Therefore I won’t comment on the modelling part (Sect. 2). Regarding the statistical aspects, a few aspects should be clarified and further elaborated before publication. Therefore, I recommend additional corrections by incorporating, if possible, the following recommendations.

1) Return level estimates

1.1. The authors stress several times in the manuscript that the estimate of the return period of the water level is ‘robust’. I agree with them that with a dataset representing more than 5000 years of cyclone activity, robustness should be achieved for estimating 100-year return levels. However, for a 500-year return level, some statistical uncertainty could still affect the result. This should be analyzed more carefully. As far as I understand the procedure, the authors calculate the empirical percentile using the results of the ensemble ('ranking-based statistical analysis' as indicated in line 321). I would expect the authors to calculate some confidence intervals, for example using bootstrap approaches; in particular, the results in Figure 8 should be further discussed in relation to this additional uncertainty estimate. An additional interest is to support the discussion in Sect. 5.3, in particular for the comparison with the study of Leijnse et al. who provides such uncertainty estimates.

1.2. A second aspect is the comparison of return levels to observed surge levels during cyclones. Could the authors consider the relevancy of using the surgedat dataset to this aim? http://surge.climate.lsu.edu/data.html

2) Population exposure

I appreciate the efforts made by the authors to discuss the limitations of their approach. In addition to the limitations raised in Sect. 5.6, could the authors also consider / discuss the use of alternatives population dataset. For instance, the Global Human Settlement Layer - Population Grid r2019a has a spatial resolution of 9 arc sec, and the WorldPop Global High Resolution Population Denominators has a spatial resolution of 3 arc sec. See references below.

Florczyk, A. J., Corbane, C., Ehrlich, D., Freire, S., Kemper, T., Maffenini, L., Melchiorri, M., Pesaresi, M., Politis, 1715 P., Schiavina, M., and others: GHSL data package 2019, 29788, 290498, 2019

Lloyd, C. T., Chamberlain, H., Kerr, D., Yetman, G., Pistolesi, L., Stevens, F. R., Gaughan, A. E., Nieves, J. J., Hornby, G., MacManus, K., Sinha, P., Bondarenko, M., Sorichetta, A., and Tatem, A. J.: Global spatio-temporally harmonised datasets for producing high-resolution gridded population distribution datasets, Big Earth Data, 3, 108– 1780 139, https://doi.org/10.1080/20964471.2019.1625151, 2019.

3) Use of JTWC dataset

3.1. As far as I understand the use of JTWC dataset for Fig. 4 is not a validation per se but the objective is to show the consistency of the ensemble results. However we note some discrepancies in Fig. 4(b) and (c) that deserve some additional comments or clarifications. In particular, the frequency for April drastically differs between JTWC and the ensemble approach. Adding some errorbars to these histograms may here also help nuancing these differences.

3.2. Is JTWC dataset used to estimate the average annual frequency of 0.70314 (indicated in line 285)? If so, please specify.

3.3. On page 19, in line 400, the authors state that ‘This landfall pattern corresponds to previous observations that the landfalling cyclones in the Bangladesh coastline tend to move north-eastward’. Is this result also confirm by the analysis of the JTWC dataset?