The research work presented in the paper “Assessing the ability of a new seamless short-range ensemble rainfall product to anticipate flash floods in the French Mediterranean area” aims at the evaluation of the added value of a seamless short range (0-6h) ensemble quantitative precipitation forecast (QPF) product, called PIAF-EPS for flash floods forecasting.

The work shows the results in terms of the skills of the system comparing the PIAF-EPS system with different deterministic QPFs calculated for eight intense rainfall events occurred between 2019 and 2021 in the French Mediterranean region.

The paper is quite-well written and readable even if some adjustment can be made in the order of the paragraph to make it clearer; the abstract and conclusions are satisfactory; the scientific methods and assumptions valid and well outlined. As a general comment I think the work of the paper is interesting from the point of view of the method of the research and for the analysis of the results. There are some parts less clear than other that need improvements and maybe further insight in some sections. Corrections and comments can be found in details in the pdf attached.

General comments: 

This article presents a comparison between different QPF-forced hydrologic predictions of flash flooding events, featuring a recently developed ensemble technique at Meteo France. Selected events are used to assess the ability of these predictions to detect the occurrence of streamflow values exceeding defined frequency-based thresholds. QPE-forced hydrologic simulations are used as reference data. Metrics based on contingency table data such as the Critical Success Index are used to assess skill. While I don’t see any new scientific insights into QPF-forced hydrologic forecasts, this is a good contribution to the literature of tools with operational implementation. The article’s methods are based on robust work previously published by others. The manuscript is written well and in a concise manner. My only concern is that the authors do not provide enough details about the featured technique that is central to the study. My recommendation is that following some minor revisions, they paper could be accepted. 

• My main comment is that the PIAF-EPS methodology is not described with enough details. The workflow schematic in Figure 2 does not provide any information on how the perturbations are computed, which seems to be an important aspect of the technique. Authors should include an example of these perturbations, so readers can get a sense of what they look like.
• Many acronyms not spelled out the first time they appear in the text. At least some of them are spelled out later in the document, but they should be spelled out as soon as they are used for the first time, so the reader is not left wondering about it.

Specific comments:

Line 61



“…for flash flood nowcasting purposes” Is it appropriate to say “flash flood nowcasting”? I have only seen nowcasting being used to describe QPE extrapolation.

Line 81

“SMASH” (L80)

Is “PANTHERE” an acronym? If so, please spell it out.

Line 94

What is “…a ten-minute observation cutoff”? Do you mean only the first ten minutes worth of observations are assimilated? How many observations (how many data points) are actually assimilated? Also, specify what data are assimilated (radar, satellite, rain gauge?).

Line 96 – 97

Spelling out this acronym (PIAF) should occur earlier in the document, as soon as it is first used. Same with all other acronyms.

Line 92 – 112

Use of “Lead time”. Consider replacing the term “Lead time” with something like forecast length, or simply referring to a particular forecast by its length. For example, the 3h forecast, to refer to a forecast that goes out 3 hours into the future. The term “Lead time” implies skill associated to a particular forecast length, and not a configurable parameter.

Line 119

With “equiprobable”, do you mean perturbations are “drawn” from a uniform probability distribution?

Line 121

What do you mean with “subrandom”?

Line 126

A better term to replace “lead time” here would be forecast length.

Line 172 – 173

How costly? How often is the system changing?

Line 176

“…the importance of hydrological reaction response”?

Table 1

“Duration” does not seem appropriate. Use “Date”? Also, I see the order of the events in here is by date, but labels “A-H” are all over the place. Not a big deal, but this is very odd order to follow here and Figure 5. It feels like the labels’ purpose was to make it easier for the events to be organized/classified, but the way they are presented in this table seems to defeat said purpose?

Line 211 – 212

The word “assimilated” is misleading. Do you recursively use reference streamflow to improve model states and/or parameters? If not, then I strongly recommend using a different term here.

Line 255

Why was the 60^th percentile chosen for the comparison? I could not see anything in the previous texts that would give indication that a particular percentile was to be used.

Line 360 – 361

More than confirming, a robust study on a large enough sample dataset should inform how truly valuable and applicable is the ensemble-based technique, particularly in real-time.