|I want to thank the authors for making extensive improvements on their manuscript. The method now feels more complete and the figures have much improved. However, I am still hesitant about some of the results and concerned for the presence of artifacts in their results. |
the revised manuscript still reads:
“In the northern Netherlands where more than 100 mm precipitation is observed” -> this is not correct. The intense rainfall was in the south(east) of the Netherlands; not the north. As can also be seen in Figure 2 in the paper.
“regions near Markermeer and IIsselmeer, and regions around Hollands Diep and Meuse River are largely affected by the flood” -> as stated in my first review round, the regions near Markermeer and Ijsselmeer were in reality not affected at all. I can’t judge where the authors base this finding on but am still afraid there may be artifacts in the methods if this is what they found with the RAPID methodology. Figure 4 shows a part of central Netherlands (not northern) with the rivers Meuse, Rhine, Ijssel, and flooding of floodplains there. This I can confirm happened last summer. But I would phrase it as such (i.e. that the floodplains of the Meuse, Rhine and Ijssel were affected); as opposed to referencing Hollands Diep, which is a permanent water body).
As mentioned in my previous review, the flooding was, by far, the most extensive in the south of the Netherlands (province of Limburg). I would challenge the authors to examine their results around the village of Meersen (confluence of Meuse and Geul) and maybe Valkenburg (along the Geul), which were both severly affected. I can imagine that flooding along the Geul river may not be picked up well as it is a small river with a limited floodplain (much more v-shaped river valley). It could be simply due to resolution issues that this is not picked up; or because the ICD module in the RAPID methodology may discard it if the Geul is not recognized as permanent water body (again due to resolution issues). It is ok if this is the case, but would really like the authors to scrutinize the RAPID methodology on where it does and does not work well. This event gives a great opportunity for that by examining known hit areas. I can imagine that it can pick up very well the flooding of floodplains along major rivers, but less so flooding in smaller upstream rivers with considerably smaller floodplains.
I am still very concerned by the results for Southern France. As mentioned by the authors, there was hardly any rainfall there (10mm). I did some searching myself and did not find any mention on floodlist for flooding in that region during that time. What further fuels my concern is the pattern of flooding that is presented in Figure 3. This looks very patchy, unlike the flooding near Louhans and along the Meuse/Rhine/Ijssel rivers (fig1 and fig 4). Given the absence of rainfall, the only flooding I can imagine happening is along the main river that may have received rainfall (far) upstream. But that should not look this patchy and I hardly see flooding linked to the main river. I did notice that also the dry reference (in blue) from June 22 looks very patchy, with many blue spots that in places where on google satellite picture I see little to no permanent water (for as far as I can judge with just visual comparison). These (in my view) questionable blue spots my result in the RAPID methodology not discarding spots, where in reality it probably should.
Something similar could be the problem with the Ijsselmeer/Markermeer area in the Netherlands which the authors referred to, but I couldn’t find the image for that region on the AWS. Btw, the Nijmegen and Roermond figures on AWS there look very good, even the smaller river of the Roer seems to be well captured!
Landsat and validation:
The authors seem to have introduced another RS source to estimate flooded areas with, using landsat-based flood inundation maps. However, this is not explained in the methodology (but introduced out of the blue in the results), nor is it clear whether this was an analysis by the authors themselves, or an independent source (there is no citation to anything). Also the reason for doing this (in this one case) is not made explicit. I can guess that it is for valiation, but as this seems to concern a comparison between two estimated flood maps (probably with similar methodology, just different input maps; but hard to judge as no details are given on the landsat based FIMs), it does not add too much value in my opinion. Much more valuable would be validation with independent empirical information such as eye-witness reports (like newspaper reports through floodlist, or social media reports through globalfloodmonitor.org).
In short: there are several results which I find questionable based on my own knowledge of the event (flooding Ijsselmeer and Markermeer), and based on the results shown by the authors (spotty pattern in SE France). Consequently, I’m afraid there may be incorrect results in certain areas. I would urge the authors to scrutinize their results in order to find the reason for this. When doing so, the authors can also reflect on when (e.g. when correct dry image is present) or where (e.g. river with substantial floodplain area) the RAPID methodology does very well, and when/where it doesn’t perform well. This would be very valuable for the rapid flood mapping community and wider risk management community to learn from.
The work presented in this Brief communication includes a concise yet accurate description of the West European flood 2021 and several features on the inundation extent, particularly in agricultural areas. Information that is to date scarcely available. Although it brings limited research novelties, the contribution is timely. Another important point is that it comes with an extensive dataset of satellite derived inundation extent both in georeferences tiff and as images. I’m in favor of publication provided that the following comments are adequately addressed:
Figure 1: Please make country borders thicker. Now they are difficult to see
Figure 2: I suggest enlarging the 4 panels on the left and reducing the one on the right. The inundation extent is more informative than the overview map.
L 83: I suggest avoiding nested parentheses “))”
L 92: “the Netherlands” I think
L109-110: This is a sentence for an Abstract or Introduction, rather than for Closing remarks
I think that the information in Figure 3 is already included with more details in Figure 4. I suggest deleting Figure 3.