This paper introduces a new method for quantifying storm hazards and their effect on building habitability, specifically in the context of Hurricane Irma. The proposed methods offer some insight into how hazard levels (e.g., water depth, flow velocity, etc.) can be used to estimate the probability of a building becoming uninhabitable following the hurricane. However, there are several missing discussions and insights. The criteria for building habitability are overly simplified and do not account for human behavior, such as voluntary displacement despite a building remaining structurally sound. The sample size is also limited—only 920 buildings were analyzed, with just 12% classified as uninhabitable—raising concerns about potential overfitting in the regression models. The model validation is less robust than expected; for example, it lacks overland flood comparisons and relies primarily on offshore water level gauges. Despite statistical significance, there is substantial overlap between habitable and uninhabitable buildings, which undermines confidence in hazard level as a strong predictor of uninhabitability. The addition of a multivariable model does not improve predictive performance, calling into question the benefit of increased complexity. Several improvements could strengthen this work: increase the sample size by including more regions, hurricanes, or buildings; revisit and expand the discussion of model limitations; clarify the assumptions behind using cell phone data as a proxy for habitability; and test whether altering the return-date threshold affects the results. Finally, the applicability of this method to regions outside the U.S. is unclear. Given the U.S.-centric dataset, it remains uncertain whether this approach could be generalized globally.

The following items show concerns of the paper itself that I think should be remedied before publication acceptance:

• Line 24 - (or hurricanes) This phrase in parenthesis makes it seem that all tropical cyclones are hurricanes when hurricanes are tropical cyclones of a certain wind speed threshold, consider rephrasing.
• Lines 23 - 26. Sentence is confusing as written
• line 26 - 'The significant loss...' should be losses
• line 28 - why are damage and fragility functions in quotes here but nowhere else?
• line 52 - household displacement 0 days after Irma implies that people evacuated? Wouldn't 1 day following the hurricane be a better metric for damage?
• lines 65-66 - Discussion of hurricane Irma should be in methods? Expand on estimated flood and storm surge volumes for the areas of interest.
• lines 69-70 - Does the cell phone data indicate that the buildings returned to their previous levels of occupation?
• line 83 - 'The extend the model is from...' Awkward phrasing
• line 85 - why was 80m chosen?
• line 87 - 150 m, why not overlap D-Flow and SWAN's resolutions?
• line 96 - spell out HEC-RAS first
• line 103 - explain what a spiderweb grid is
• line 109 - how was SWAN drag coefficient determined to be insufficient?
• lines 109 - 115 - explain more clearly the methods on updating the drag profile and wind field values
• lines 174-176 - '...essential services such as power and schools were recovered by this point.' What point, when was this data obtained, what date and how long after landfall?
• lines 178 - 183 - confusing paragraph, rework
• lines 186 - 192 - this is methods
• lines 316 - 320 - This discussion is what I was waiting for the moment I first saw figure 4, can this be addressed earlier?
• model validation - can the model be further validated with flow depths, not just offshore gauges?
• Figure 4 - is this for all buildings in both counties? Please expand in figure caption, including the binary outcomes at the top and bottom axes makes this style of figure unconvincing of the habitability functions being useful tools, the fact that there are uninhabitable buildings at low or 0 depths, flow speeds or sig. wave heights make me think that the probabilities are suspect.
• All figures in the style of figure 4 should have a legend, the captions should not make the reader turn back to figure 4 to remember what each lines means.
• Figure axes and ticklabels are very small, consider enlarging for the average user
• Figure 5, where is the confidence interval reasonable, because teh dotted lines are only close for moderate values of the hazard included. - see notes on figure 4's binary outcomes
• Figure 6 - are we still only looking at the county that has building data or is this all data again?

The manuscript outlines a methodology for defining habitability functions, which are purported to be a more accurate reflection of the impact of natural hazards and the ability of a community to recover than earlier work on damage or fragility functions. The authors focus on the impact of Hurricane Irma on locations on the Atlantic coast of Florida. The method couples information from a hurricane surge model (Delft3D-FM) with information from location based services and property data to deduce if residents of damaged buildings have resumed normal routines, linking this deduction to habitability of their homes. They indicated that the impact of water depth (flood depth) appears to be the major influence on habitability, greater than wave height or water velocities. 

I think this is very interesting, perceptive work. I do have a few comments: 

1) Figure 1 shows the model grid. This seems very small for hurricanes. The grid implies the assumption that water level changes generated outside the grid due to the hurricane are negligible. This may or may not be the case for this specific storm event, but is not generally the case, as many hurricane researchers using ADCIRC use their standard grid, which covers half of the Atlantic Ocean. There has been work that suggests that a small grid might miss surge forerunners and other possible motions that can cause additional damage aside from the main surge event. This might explain why the model is incapable of simulating the long surge buildup (Figure 2d). If the emphasis is on peak surge, then perhaps it doesn't matter, as the model seems to be sufficiently tuned to get the max surge right. But to what degree does this impact the velocities (and, in turn, impact the finding that the habitability functions developed with velocities perform poorly)? 

2) Also regarding modeling: the tidal conditions from Egbert and Erofeeva can be less accurate in very shallow water, such as that near the Bahamas, where the offshore boundary is located. Was this accounted for?

3) While I understand the presumed connection between habitability and the resumption of a normal routine originating from the same dwelling, the definition of "habitability" might be somewhat ambiguous. After Katrina, many residents lived in their homes while being compelled to return to their routines, yet many of these homes had no power or water. These homes served as functional shelters but that shouldn't be confused with recovery, since they were far from recovered. In many cases, these residents were out of options. This might actually bias the reliability of these habitability functions against those with lower incomes and fewer options. I guess I would like to either see a clearer definition of "habitability" (i.e. dwelling with sufficient cover from the elements), or these habitability functions placed in a more general context. 

4) The method relies on the availability and amount of LBS data. Collier and Monroe Counties have almost half a million residents between them. What would be possible in a place like the Louisiana coast? St Mary Parish, the largest coastal parish, has barely 50,000 residents. Is there a lower limit on data which would make the method meaningless?