Climate Change Flood Risk Analysis: Application of Dynamical Downscaling and Hydrological Modeling
Round 1
Reviewer 1 Report
The study is interesting but lack of novelty. The literature review is not complete. "Water, Air & Soil" published a similar study using HEC models with CMIP5 climate model results for analyzing climate change impact on flood . See " Assessing the impact of climate change on flood events using HEC-HMS and CMIP5" on the journal. With this being said, why only RCP8.5 is used in this study? Justification will be needed to the minimum.
The selection of GCM and hydrologic models have significant impact on results. There are many publication such as by David Hill recently discussing this. If you only use one GCM model and one hydrologic model, acknowledge of the shortcoming will be needed at least.
ok.
Author Response
We would like to thank the reviewers and the associate editor for their valuable comments and suggestions, which have helped to improve the manuscript. Thus, this revised version was prepared taking into consideration all the comments of the reviewers.
What follows are responses and clarifications for reviewers’ comments.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
GENERAL COMMENTS
The study represents a flood risk analysis for a small (26.5 km2) river basin in eastern Brazil for the near past (1979-2018) and near future (2019-2048), combining hydrological modelling with a climate change projection generated by a high-resolution (5 km) regional climate model. Peak flows resulting from maximum 6-hourly rainfall in near past and future climates are simulated for several different scenarios regarding the distribution of the rainfall in the basin. In nearly all cases, the peak flow exceeds the current capacity of the drainage system. Climate change is found to moderate the maximum precipitation and peak flows up to the 30-year return period, but very extreme events (return period 50 to 100 years) increase in magnitude.
The study is a good example on the use of high-resolution, bias-corrected climate model output in the generation of hydrological scenarios. The methods used in the generation of the scenarios and the data analysis make good sense. However, there is an important limitation: only one climate change projection is used. This means that the results must be quite uncertain, but the uncertainty is difficult to quantify because there is only one simulation. This issue is exacerbated by the fact that the analysis focuses on changes in the extreme upper tail of the precipitation (and peak flow) distribution, where the statistics have a very low signal-to-noise ratio. As there are only 40 years of model data for the present past and 30 years for the near future, the estimates for multidecadal return values of rainfall may be drastically affected by the most extreme individual precipitation events in the simulations.
The use of only one climate projection is a pragmatic choice: this is the only simulation that is available for the authors and has so high resolution (5 km) that it can (at least in theory) simulate precipitation on the scale of the studied river basin. Yet, even if one accepts as a fact that only suitable model is available, one may ask whether it is optimal to use data from just one grid box of the model. Climate models are not expected to be skilful at their smallest resolved scale, and it is very unlikely that they can reliably tell how the change in extreme precipitation in one grid box will differ from the grid boxes in its neighbourhood. Therefore, one way to partially assess the uncertainty in the results would be to repeat the calculations using (instead of the grid box right over the basin) precipitation data from other climate model grid boxes from its vicinity, and to study how sensitive the results are to this choice. Because bias correction is used when applying the climate model output in the hydrological model, such a “violation of reality” would most likely not have any large impact on the quality of the hydrological simulations.
I realize that the suggestion outlined in the previous paragraph may be laborious to implement, and therefore leave it to the authors to decide how to proceed. At the very minimum, however, the uncertainty that results from the use of a single grid point for a single climate model simulation should be explicitly discussed in the paper.
Aside from the issue discussed above, most of my remaining comments are minor suggestions for improving the presentation.
DETAILED COMMENTS (everything except for language)
1. L21. 1.79 % means one simulation out of 56. Thus, two-decimal accuracy is not justified.
2. Figure 1. The combination of red and green colour should be avoided as a courtesy to those with colour blindness.
3. L112 and L123. Daily or 6-hourly rainfall?
4. L155-161. Please report the absolute percent coverage of the different land use classes in each of the years 2005, 2016 and 2019 (possibly as a table). Otherwise, it is difficult to know how important, for example, a 140 % increase in exposed surface is.
5. Figure 4. The maps for the years 2005, 2016 and 2019 look identical.
6. Section 3.2 (Catchment subdivisions) should rather be placed within Section 2.1 (Study area) and not in Section 3 (Results).
7. L214-215. How large is the peak flow required for channel overflow?
8. Table 1. The unit of peak flow should be m3 / s.
9. Table 1. Why are the rainfall values largest for the Spring and smallest for the Low part of the catchment? Is this just because the return values of rainfall decrease with increasing sub-catchment area? Also, why is the rainfall in the Medium and Low parts of the catchments smaller in the simulations where only these catchments receive rainfall (TR05 M and TR05S) than in the simulation (TR05) in which all catchments receive rainfall? See also the next comment.
10. L231-233. The rainfall values reported for TR05 L in Table 1 are much lower than those given here. Why?
11. L237-240. I assume that this difference results from a low bias of the simulated 5-year return values of precipitation. Please note this explicitly in the text.
12. L242-243. Based on Table 2, there is an underestimate for all return periods, not only for those below 50 years.
13. L251-256. The text on these lines can be written more compactly in one sentence: “Table 2 reveals a decrease of the return values of 6-hour precipitation from the present climate simulation to the future climate simulation for return periods up to 30 years, but an increase for the 50-year (5.43 %) and 100-year (12.52 %) return periods.”
14. L256-259. This behaviour, i.e. an increase in precipitation at very high intensities even where there is a decrease at lower intensities, is qualitatively consistent what is more generally seen in climate model simulations (e.g., IPCC Working Group 1 AR6, Chapter 11). On the other hand, with only 30 years of simulation data, the changes in multidecadal return values of precipitation are subject to very large sampling uncertainty and one should therefore be cautious with any conclusions related to them. See also the general comments.
15. L279-281. The general understanding is that the distribution of precipitation intensities will shift towards higher values in a warming climate, meaning a decrease in the frequency of precipitation in most areas but an increase in the highest intensities nearly everywhere (e.g., IPCC Working Group 1 AR6, Chapter 11). However, the changes are regionally variable, and one should therefore avoid generalizing the results of regional studies to other areas. Furthermore, the shift from decrease to increase typically occurs on much smaller intensities than that corresponding to the 50-year return value. Thus, your model results may be somewhat atypical, either due to the model used or because of internal variability in the simulations.
DETAILED COMMENTS (language)
16. L20. surface runoff and evaporation?
17. L29-30. high level of imperviousness?
18. L57. was developed / was first applied
19. L80. shorter: "evaluated the performance"
20. L90. What does "effectiveness" mean in this context? Could you use a more explicit wording?
21. L115-116. Do you mean "total annual precipitation" or "annual maximum 6-hourly precipitation"?
22. L169. "negatively interferes with" or simply "reduces"
23. L170. surface runoff and evaporation
24. L176. Figure 5 shows the subdivision proposed in this study.
25. L194. closest to the study area?
26. L198. was developed
27. L201. Quantile plot of observed annual maximum 6-hour precipitation with Gamma distribution fit?
28. L207-208. shows the studied combinations of return period and spatial distribution of rainfall
29. L218. precipitation has
30. L223-224. peak flow ... occurs
31. L226. Peak flow and rainfall statistics for 5-year return period.
32. Table 1. Suggested changes in wording / terminology: (i) Flow bias percentage à percent difference in flow; (ii) rainfall bias percentage à percent difference in rainfall. Calling simulated differences between future and present climates "biases" is confusing.
33. L267-268. The climate model simulated lower than observed precipitation extremes in the present climate.
34. L284-285. model underestimated the intensity of precipitation extremes, despite the application of a bias correction.
35. L290-291. that downscaling with Eta/HadGEM2-ES simulations results in good agreement with observations of mean annual precipitation?
The language is mostly OK but there are still some grammatic and wording errors, as well as some unnecessarily complicated sentences. See Detailed comments 16-35 for sugestions for improvement.
Author Response
We would like to thank the reviewers and the associate editor for their valuable comments and suggestions, which have helped to improve the manuscript. Thus, this revised version was prepared taking into consideration all the comments of the reviewers.
What follows are responses and clarifications for reviewers’ comments.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I have no additional comments.
