Dear authors,

Thank you for your submission to HESS. We have received two valuable reviews from the referees. Reviewer #1 is very critical, and he has provided very valuable comments that you have mostly taken on board and answered in your rebuttal letter. Reviewer #2 is more positive and has also provided valuable comments.

I tend to agree with reviewer #2 in that this contribution is potentially valuable, but I can understand the concerns of reviewer #1 whose comments are extremely useful and need to be appropriately addressed. I can see that you have welcomed most comments and have done extensive work to address them, but I also believe that there is still some work to be done (as explained below).

Though I understand the concern of reviewer #1 regarding the estimation of recurrence intervals that can be compromised by the regionalization technique used in the study, I tend to agree with the authors in that the use of a regionalization technique (i.e., space for time) that pools data from neighbouring basins can be useful to construct an extended record. This type of approach can potentially be the only way to extend the rainfall or discharge time series for frequency analysis in areas with short precipitation records, and has therefore been used in a number of published studies (in many cases for practical applications, see for example Soong et al., 2004; among other studies). It will be useful to appropriately acknowledge and briefly mention some of the previous research (using space for time approaches) in the paper, establishing similarities and differences to these previous procedures. The limitations of applying this type of technique need to be also appropriately acknowledged (see also Soong et al., 2004, Stedinger et al. 1992). For example, one of the assumptions of the index-flood procedure (that uses regional data) is that observations at different sites must be independent (i.e., not subjected to the same rainfall event). However, as discussed in previous work, this assumption is unlikely to be satisfied for frontal rainfall events as adjacent sites can be correlated (Hosking and Wallis, 2005). But even if this is a limitation, Hosking and Wallis (1988) showed regional frequency analysis can (in some cases) be more accurate than at-site analysis even when inter-site (spatial) dependence is present. You can probably add these or alternative references.

According to the rebuttal letter (in the response to reviewer #1), you have now modified (in the revision) the methodology that you use for the frequency analysis to include the “Peak Over Threshold” (POT) analysis. I find this explanation in the rebuttal letter unclear (and even less clear in the revised paper). The POT method is based on utilizing all peak events of a given time series exceeding a specified threshold. The selection of the threshold value in the POT method is not straight-forward, and you are not providing enough information of this analysis. You seem to be using a threshold of zero (?); this means that you are using all the data? By considering several peak events in each year instead of annual maxima, the number of data points for statistical processing increases (which is the whole idea of the POT, Wallis et al., 2007). However, the basic assumption of the POT method is that the threshold is high relative to typical values, and a lowering of the threshold to increase the data may lead to bias (see for example results for varying thresholds in Davison and Smith, 1990). This needs to be clarified; the explanation in the rebuttal letter (and paper) does not seem to agree with the use of POT. If POT is not being used, then please make clear how are you selecting values to calculate ARIs for rainfall (the maximum annual?).
How is the threshold helping avoid double-counting?

In addition, I find that the methods in the paper are still unclear: what is the probability function used in the analysis, are you using the empirical values directly? Please explain and justify if needed.

Please note that the response to Q10 of reviewer 1 is still unclear. Please clarify the methodology used (and its rationale in terms of exceedance probability calculation of rainfall/discharge) both in the response to this reviewer and in the revised manuscript.

I wonder if there are any available rainfall records for at least one the catchments that you can use to compare the results of the ARI (annual recurrence interval) obtained from the radar (even if just for that catchment).

As mentioned above, the methods in the original paper submitted for discussionbwere not well explained. You have partially addressed this issue but there is still a need to further clarify the points stated above. Perhaps a general schematic or a summary of methodological steps could improve the clarity of the methodological approach.

I look forward to receiving the revised manuscript and new response. Please note that the paper and rebuttal letters will be sent back to the reviewers.

Sincerely,
Patricia Saco
Editor, HESS


References:
Davison, A. C. and R. L. Smith (1990). Models of Exceedances Over High Thresholds. Journal of the Royal Statistical Society 52 (Ser. B), 393–442.
Hosking, J. R. M., and J.R. Wallis. 2005. Regional frequency analysis: an approach based on L-moments. Cambridge: Cambridge University Press.

Hosking, J. R. M., J. R. Wallis, and E. F. Wood. 1985. Estimation of the generalized extreme-value distribution by the method of probability-weighted moments. Technometrics, 27 (3): 251-261.

Soong, Audrey L. Ishii, Jennifer B. Sharpe, and Charles F. Avery, 2004, Estimating Flood-Peak Discharge Magnitudes and Frequencies for Rural Streams in Illinois, USGS Scientific Investigations Report 2004-5103.

Stedinger J. R, R. M. Vogel, and E. Foufula-Georgiou (1993) “Frequency analysis of extreme events” in Handbook of Hydrology, R. Maidment, Ed., chapter 18, pp. 18. 1–18. 66, McGraw-Hill, New York, NY, USA.

Wallis, Schaefer, Barker, and Taylor (2007) Regional precipitation-frequency analysis and spatial mapping fpr 24-hour and 2-hour durations for Washington State, Hydrol. Earth Syst. Sci, 11(1), 415-442.

I agree with the reviewer that the new manuscript addresses most of the issues raised in the previous reports. I also agree that it will be very useful to the readers to add to the conclusions section an overview of the main limitations, i.e. linked to the general applicability of the proposed methodology, and a clear statement of the main underlying assumptions. These are just minor revisions that will further improve the paper.
I look forward to receiving your new manuscript.
Best Regards,
Patricia Saco

Dear authors,
Thank you for addressing the reviewer's concerns and submitting an improved version of the manuscript.
Regards,
Patricia Saco