|The study reports an interesting point of view to be considered in practical applications of flood damage assessment. The manuscript is well written, and the methodology, data and results are well discussed, however it is very case-study specific. So, I suggest a few elements that should be improved to make the work more useful to support the application of damage estimation models in other sites.|
As from figure 8, table 5 and comments in the text, it is clear that all the observed and estimated damage are not correlated. This fact is very interesting and deserve to be investigated in more detail. How these differences are located in space? Is there any spatial pattern of model overestimation and, are these patterns dependent on topographic/hydraulic parameters?
The “transferability” of a model is recalled in different points of the text. However, the study does not give a real support to the transferability because only a single case study is analyzed. Instead, “internal transferability” could be useful and should be discussed in more detail: I would have expected that the models were also calibrated (or parameterized on the basis of a blinded dataset) only on a subset of the 345 known building and then validated on the remaining complementary subset. Results from this application could then be used to “transfer” the application to the set of (877-345) “ungauged” buildings where no damage data is available. Do the authors have tried this option?
Page 7, line 166 and throughout the text. It is not clear if each research group have implemented all the selected models (and there are multiple realizations of the same model) or if each research group have implemented one model.
Page 7, step 4. This step is commonly referred to as sensitivity analysis; the authors may evaluate if recalling this definition can improve readability.
Page 21, line 442. Delete “in, as.”