Dear authors,

Thank you for letting me read the interesting manuscript that provides valuable results on urban flood risk scenarios for Shanghai. The manuscript fits the aims and scope of the Journal Natural Hazards and Earth System Sciences. The authors present their method and data transparently as well as provide a good presentation of the results. However, the manuscript lacks a general discussion of studies on urban flood risk assessment and a link of their findings to the literature. Therefore, I recommend the authors to add this to the manuscript. Also, I suggest a external language check. The manuscript includes some grammar mistakes. Next to smaller remarks I therefore recommend a major revision.

Minor recommendations:

Please do not use abbreviations, such GE or BU in the abstract.

Please rephrase: „We also find that urban will tend to expand to areas vulnerable..“  What is meant by urban? Urbanization?

What is meant by „coupling model“ (line 24)?

Lines 31-32: „The United Nations reports that the global population will increase by 29% (7.6 billion) between 2017 and 2050 (United Nations, 2017b),“ Is the increase taking place in coastal cities? If yes, please make this clear in the sentence.

Line 44: Please give some examples what is meant by „environmental factors“

Line 54: „The FLUS model improves the simulation accuracy of the model…“ Which model will be improved?

Please rephrase: „To answer this question, we first consider how urban grow under different environmental and planning factors in the future. (lines 61-62)

Please include a presentation of the further content of the paper in the Introduction.

Line 149-150: Can you please further explain why do you choose 2,768 km2 in 2030 and 3,200 km2 in 2050 as reasonable city growth pathways?

Line 169-170: Please add some references to justify the useful applicability of the LISFLOOD-FP model.

Line 187: Please explain what is meant by the abbreviation LUCC.

I propose to include the explanations contained in the titles of the figures and tables in the text of the manuscript and to refer to the figures and tables.

Lines 250-253:  I suggest presenting either relative numbers (in %) or the absolute number of inundated land area. For RCP2.6 scenario you are using the land area but for RCP 8.5 you present relative numbers.

Line 260-261: Please make clear whether this findings means in absolute and/or in relative numbers.

Line 329: „…range and spatial distribution of flood risk in future urban“ Please add „areas“ at the end of the sentence.

The paper tackles an important subject in future management perspectives of coastal urban regions. Nevertheless, there are some major constraints/critisism to the current version of the paper:

- the land cover data set seems to be outdated (2014) and has only a coarse spatial resolution (100 m). Based on this, to model the year 2015 with land cover data from 2014 is not so difficult and shows a high coincidence as expected. Why not using a higher spatial resolution from Sentinel-2 (10 m) from the year 2020? Six years after the acquired land use data set from 2014 this would show whether the performance of the land use model is good enough or not...,

- the ASTER-DEM used has a 30 m spatial grid, but everybody knows, that the vertical accuracy may vary up to 5 m and more. This is a major drawback in coastal lowlands, where just small height differences may cause large discrepancies in flooded areas. Better use LIDAR data if available,

- the single forward modelling of the urban development may not concider the polycentric development of the agglomeration,

- land subsidence is not equally everywhere. It depends very much on the ground substrate. Fluvial sediments may subside more than rocky underground. And it also depends on the anthropogenic use. Roads on sedimentary ground, where heavy trucks are driving each day may subside much more than anywhere else... It's not enough to analyze that statistically, one would hove to look attentive where this would happen. Interferometric evaluation of multitemporal microwave data would provide a propoer estimation on that...,

- the result showed a stronger inundation by the GE scenario than with the GP scenario. Flooding per se is not bad, so one would not rank the GE scenarion worse than the GP scenario, since it might consider clean air allies or urban green spaces. The authors mention that in their chapter 5.2 "Recommendations", but should emphasize that much more...