Testing Urban Flood Mapping Approaches from Satellite and In-Situ Data Collected during 2017 and 2019 Events in Eastern Canada
Round 1
Reviewer 1 Report
This paper tested multiple urban flood mapping methods by using different data sources, including RADARSAT-2, Sentinel-1, Planet, RapidEye, DEM, hydrometric station, citizen geotagged report, and oblique pictures. The research topic is important and the method is standard. The flood mapping methods are previously published so this paper did not focus on providing novelty. The comparison of methods is interesting for readers in this field.
The paper could be revised to improve its readability. First, reduce the length. The manuscript includes too much general information. Unless the authors think this is for general audiences, the current version is too lengthy. Second, provide an overall summary of the different data sources, such as table including accuracy comparison. Important findings and their positions in the current literature should be highlighted. Third, provide a data-sharing plan. The case studies may not be enough to draw decisions. Sharing the data can involve more cooperations thus citations.
Author Response
Reviewer #1
This paper tested multiple urban flood mapping methods by using different data sources, including RADARSAT-2, Sentinel-1, Planet, RapidEye, DEM, hydrometric station, citizen geotagged report, and oblique pictures. The research topic is important and the method is standard. The flood mapping methods are previously published so this paper did not focus on providing novelty. The comparison of methods is interesting for readers in this field.
The paper could be revised to improve its readability. First, reduce the length. The manuscript includes too much general information. Unless the authors think this is for general audiences, the current version is too lengthy. Second, provide an overall summary of the different data sources, such as table including accuracy comparison. Important findings and their positions in the current literature should be highlighted. Third, provide a data-sharing plan. The case studies may not be enough to draw decisions. Sharing the data can involve more cooperations thus citations.
Thank you for your positive comments. We agree that the paper is long and that a lot of general information is presented in the introduction. However, we felt that while presenting the background and context of these events was perhaps not absolutely necessary to a strictly scientific audience, it strongly reinforces magnitude of these events and therefore the need for this and other work on urban flood mapping that may not be obvious to some readers. Also, excluding CCD and perhaps a few other elements, the paper is intended to be accessible to a wider audience since it presents fairly simple solutions to urban flood mapping, requiring only a DEM and point-based flood perimeter observation input to a 1-D flood-filling algorithm that itself is fairly simple to code in an open-sourced language such as R. Therefore, remote sensing expertise is not absolutely necessary to implement many of the proposed methods. After a careful re-read, we agree that some of the background information was unnecessary and didn’t add much to the paper and therefore some of this material was removed. The paper remains fairly long due to the fact that a large amount of disparate data were used in numerous case studies.
A summary table is included at the beginning of the discussion section with data, methods, accuracy and validation source. Our assessments are a mix of quantitative and qualitative based on the data we had available to us. The discussion highlights what worked well, what worked less well and what didn’t. In looking into the current literature on CCD for urban flood mapping, some authors report good overall agreement with coincident optical imagery in a qualitative way (Chini), others report high overall accuracy combining coherence and intensity (e.g. Li et al., 2019). For CCD, we now include comparisons with Chini, Li et al., 2019 and Chaabani et al., 2018 on common metrics. Few comparisons were available for the use of optical remote sensing for urban flood mapping due to factors already mentioned in the paper as well cost. We now include comparisons with Feng et al., 2015 (UAV) and Jakovljevic et al., 2019 on urban flood mapping and general waterbody extraction using optical data.
A data sharing plan is in the works but is complex given all the data we used and data sharing restrictions with the Charter, RS2 etc., and is therefore currently beyond the scope of this paper.
Reviewer 2 Report
A rewarding read, thank you. I would not suggest any major changes, only clarifications in some areas as follows:
Line 143 – “Urban flood extents can be modelled with …” – some examples/ references where this has been done might be useful for this claim.
Line 154 - “Reported errors from different studies …" – reference please.
lines 153 and 118 - ‘several’ or ‘numerous’ studies stated, but only one reference given.
Fig 2. Y axis label would be useful … Would also be interested to know the median is derived from how many years of data?
Line 244 – for readers in different countries, it could be useful to be clear if your LiDAR DTM is processed so that vegetation and man-made features have been removed. Confusingly, DEMs and DTMs (versus DSMs) can mean different things, depending where in the world you are.
Line 279 – Do you have a reference for the EGS’s CGI that you used?
Line 301 – … and the OCR that you used – do you have a reference?
Line 308 – suggest should abbreviate DEM earlier in paper as already used ‘DEM’ a few times before this point.
Line 332 – Perhaps, for the general reader, you may wish to explain quickly what DN is, why you use this instead of backscatter in dB, and possibly the relationship to reflectance.
Lines 344-355 / 393 / Fig 11 – regarding validation of a maximum flood extent map with the NASP photos/video images - for repeatability - can the authors please provide a sentence to detail of how you compared them? E.g. algorithm or visual inspection?
Line 367-369 – sentence beginning “Water heights along urban shorelines …" - maybe consider rephrasing it, as is slightly confusing wording.
Fig 4 - Suggest deleting axis numbering in Fig 4, boxes 1-3 as text is too small to read.
e.g. Table 3 and Line 457 - inconsistency in word Dike/Dyke throughout manuscript.
Line 482 and Table 4 .– The 5 x 5 window units – presumably this is pixels? Also Line 647 – 3 x 3.
Fig 11 vs Fig 8. Consider using consistent colors for ‘NASP flood’ points?
Lines 280, 721 – Flood not food.
Line 749 – Is it worth clarifying that Sentinel-1 is a constellation of 2 satellites and together they have a 12 day repeat orbit in your region, but this repeat could be shorter elsewhere on Earth?
Author Response
Reviewer #2
A rewarding read, thank you. I would not suggest any major changes, only clarifications in some areas as follows:
Thank you for your positive comments.
Line 143 – “Urban flood extents can be modelled with …” – some examples/ references where this has been done might be useful for this claim.
A reference to Wagner and Lambert, 2007 was added.
Line 154 - “Reported errors from different studies …" – reference please.
A reference is provided to the Schumann paper from 2011, which was cited in the previous sentence. This paper provides a table of previous studies reporting water height errors from fusing remotely-sensed flood extents with DEM data.
lines 153 and 118 - ‘several’ or ‘numerous’ studies stated, but only one reference given.
Both instances cite review papers that include several studies. I don’t believe several references are warranted when citing review papers.
Fig 2. Y axis label would be useful … Would also be interested to know the median is derived from how many years of data?
Y-axis was added. This is now stated as ‘calculated the median and maximum for the available period for each station’
Line 244 – for readers in different countries, it could be useful to be clear if your LiDAR DTM is processed so that vegetation and man-made features have been removed. Confusingly, DEMs and DTMs (versus DSMs) can mean different things, depending where in the world you are.
We now state that the HRDEM dataset represents elevation ‘of the terrain surface’.
Line 279 – Do you have a reference for the EGS’s CGI that you used?
Now added to the references:
Decker, V.; Marquis, G. Validating EO derived products during natural disasters with Crowdsourced Geographic Information (CGI). 38th Canadian Symposium on Remote Sensing (CSRS). June 20-22, 2017, Montreal, Quebec, Canada.
Line 301 – … and the OCR that you used – do you have a reference?
Now added to the references:
R. Smith, "An Overview of the Tesseract OCR Engine," Ninth International Conference on Document Analysis and Recognition (ICDAR 2007), Parana, 2007, pp. 629-633, doi: 10.1109/ICDAR.2007.4376991.
Line 308 – suggest should abbreviate DEM earlier in paper as already used ‘DEM’ a few times before this point.
The DEM acronym is now defined in the abstract the first time it’s stated, and removed on line 308.
Line 332 – Perhaps, for the general reader, you may wish to explain quickly what DN is, why you use this instead of backscatter in dB, and possibly the relationship to reflectance.
DNs represent scaled surface reflectance. Now stated ‘NIR dark threshold values ranged between 800 and 1500 Digital Numbers (DN = surface reflectance x 1000) or (0.8 to 1.5 % reflectance), sieve sizes…’
Lines 344-355 / 393 / Fig 11 – regarding validation of a maximum flood extent map with the NASP photos/video images - for repeatability - can the authors please provide a sentence to detail of how you compared them? E.g. algorithm or visual inspection?
We now clarify that reference data are based on visual interpretation of NASP images.
An assessment of the maximum flood extent map was conducted by visually interpreting NASP oblique images and still frames captured by video in the Pontiac region just east of Ottawa / Gatineau.
After flood filling adjacent urban areas from shoreline elevations derived from RS2 flood maps, a thorough examination was conducted against known flooded urban areas based on visual interpretation of NASP images and other reference data acquired on corresponding dates.
Figure 11. Validation of simulated flood extents from single flood perimeter observations combined with Lidar DEM data along roads in each of the three boroughs, using visual interpretation of NASP, high-resolution optical imagery and the 2019 maximum flood extent product.
Line 367-369 – sentence beginning “Water heights along urban shorelines …" - maybe consider rephrasing it, as is slightly confusing wording.
We removed ‘water heights along’ and simply state that urban shorelines were removed.
Fig 4 - Suggest deleting axis numbering in Fig 4, boxes 1-3 as text is too small to read.
Axis numbers were deleted.
e.g. Table 3 and Line 457 - inconsistency in word Dike/Dyke throughout manuscript.
‘Dyke’ was replaced with ‘dike’ throughout.
Line 482 and Table 4 .– The 5 x 5 window units – presumably this is pixels? Also Line 647 – 3 x 3.
‘Pixels’ was added in each instance and in the column heading in Table 4.
Fig 11 vs Fig 8. Consider using consistent colors for ‘NASP flood’ points?
NASP point colors were changed in Figure 8 to be consistent with other Figures.
Lines 280, 721 – Flood not food.
Corrected.
Line 749 – Is it worth clarifying that Sentinel-1 is a constellation of 2 satellites and together they have a 12 day repeat orbit in your region, but this repeat could be shorter elsewhere on Earth?We now state that the S-1 constellation consists of two satellites on Line 208 in the data section.
Exact repeat data are needed for InSAR and this is available every 6 days with the constellation of two satellites. However, in our region, for reasons unknown to us currently (although we suspect it had to do with a polarization change due to RCM transponder calibration), we mainly received S1A data and little to no S1B. We’ve changed 12-day repeat cycle which was available to us, to a 6-day repeat since both satellite data are available to us going forward.
Reviewer 3 Report
This paper presents a method to extract urban flooding from a number of satellite images, including Sentinel-1 and Radarsat-2, using DEM and water level.
In general, the paper is interesting, in particular because it describes the use of R2 images for urban flooding, which is not very well represented in the scientific literature on urban flooding.
The paper is well structured and well written. I have only really one concern that in my opinion the authors need to address before publication. The methods they present is very similar to the works by Mason et al. on urban flooding using TerraSAR-X. See for instance the following paper and references therein:
D.C. Mason, L. Giustarini, J. Garcia-Pintado, H.L. Cloke,
Detection of flooded urban areas in high resolution Synthetic Aperture Radar images using double scattering,
International Journal of Applied Earth Observation and Geoinformation,
Volume 28,
2014,
Pages 150-159,
ISSN 0303-2434,
https://doi.org/10.1016/j.jag.2013.12.002.
(http://www.sciencedirect.com/science/article/pii/S0303243413001700)
The authors have not cited these works but use very similar methods. I suggest in their methods section, they include a paragraph that relates their current work to the works by Mason et al. and more so, explains what the differences and novelty of the methods presented here is.
Author Response
Reviewer #3
This paper presents a method to extract urban flooding from a number of satellite images, including Sentinel-1 and Radarsat-2, using DEM and water level. In general, the paper is interesting, in particular because it describes the use of R2 images for urban flooding, which is not very well represented in the scientific literature on urban flooding.
The paper is well structured and well written. I have only really one concern that in my opinion the authors need to address before publication. The methods they present is very similar to the works by Mason et al. on urban flooding using TerraSAR-X. See for instance the following paper and references therein:
D.C. Mason, L. Giustarini, J. Garcia-Pintado, H.L. Cloke,
Detection of flooded urban areas in high resolution Synthetic Aperture Radar images using double scattering,
International Journal of Applied Earth Observation and Geoinformation,
Volume 28,
2014,
Pages 150-159,
ISSN 0303-2434,
https://doi.org/10.1016/j.jag.2013.12.002.
(http://www.sciencedirect.com/science/article/pii/S0303243413001700)
The authors have not cited these works but use very similar methods. I suggest in their methods section, they include a paragraph that relates their current work to the works by Mason et al. and more so, explains what the differences and novelty of the methods presented here is.
Thank you for brining Mason’s work to our attention. We added the following paragraph:
This approach is similar to methods developed by [Mason], who refine urban flood extents using water heights of nearby rural areas determined by intersecting flood boundaries with a Lidar DEM. Their approach is more sophisticated than ours, since they first map urban water in unmasked regions of non-shadow or layover areas determined by a SAR simulator. An initial rural classification is used to determine backscatter and water height threshold inputs needed to classify adjacent urban areas. Dark pixels in unmasked urban regions assigned as water based on the previously determined backscatter threshold then serve as seeds for subsequent region growing into 'unseen' areas using a water height threshold and Lidar DEM. An important difference in our approach is that we rely entirely on Lidar elevation to map urban flooding from the shoreline, whereas [Mason] use a combination of backscatter and elevation. Their method requires additional processing to run a SAR simulator to identify shadow and layover, potentially increasing the latency between data reception and product dissemination.
