Dear Authors,
the topic is extremely interesting and timely in order to help obtaining information on flood risk for many poorly gaged parts of the world.

The weakest point of the work is the comparison with the remote sensing estimates, that is, instead, crucial in order to understand how much we may rely on the reconstructions made by people, even if the remote sensing estimates are also affected by very large uncertainties (and such uncertainty is an issue that you should try to address better and more quantitatively than you do at ll. 255-260 and in particular you should refer to the limits of the specific images and methods you use for each flood event: acquisition times, resolution, etc).
The section on remote sensing (3.1.3) is to be completely rewritten, as both Referees require, providing much more information on the input data and a clear layout of the steps used to process the images and retrieving the estimation of the flooded areas. Some of the phrases you suggest in the reply to Ref2 are not clear (“Various image improvement and correction techniques have been applied” and “Since both methods are equal, the second method is used.”???).
In addition you do not explain why you use different remote sensing data for the different floods (2005 with Spot and 2009 and 2012 with Google Earth). And you should not change the remote sensing source, for a fair comparison you should use the same method for all the events. Actually, it would be better to use always the Google Earth images, since they are available in many other parts of the world and the method to estimate the flooded areas from such images, if well explained, may be reproduced elsewhere.

Other comments:

Caption of Fig. 2: specify that the flowchart refers to citizen-based methods only. And add a flowchart with the steps for remote sensing method (see comment above)

Section 3.1: Please describe better how the ‘local representatives’ are identified by the agencies. It is not clear which kind of people they are (representative of what?) and why we expect differences in respect to the chiefs (different education? age?).

On Ref2 comment:
“2/ p.5, stage 2: using their methodology, the authors obtain two information about waterlevels: the first one from the mapping after the training, the second one after the field survey. How did they resolve possible conflicting results? Which source of information did they consider as the most reliable?”
Please explain also in the revised text how conflicting results are managed.

Figure 4 and Table 3: add also the amount of overlapping between the estimates of flooded areas obtained through citizens and through remote sensing.

Dear Contact Author, dear Editorial Office,
before sending the revised paper to at least one Referee, I would ask you to amend the reply 'package':
1) addressing in the reply letter also all the referees' comments, not only mine. You may of course copy part of the replies you uploaded in the discussion, but you need to refer, in the reply letter, to the new version of the manuscript, saying where you made the suggested changes.
2) removing from the track change file all the revisions that are only on formatting or small language adjustments, so that only the substantial changes are highlighted.
3) if possibile, recapping in a table all the main changes in the revised version, specifying in which pages and lines of the revised manuscript you made such changes and which comment(s) prompted them.
All the best,
Elena Toth

Dear Authors, I fully agree with the referee that the presentation is indeed much improved.
In addition to the technical corrections suiggested by the referee, I would also suggest to clarify why you have not used the Google images for the first event, explaining also that, if possible, it would be better using the same remote sensing source.
Many thanks for having submitted ypur interesting work to HESS.
Best regards,
Elena