Please, see attached document for details.

Review: “Assimilating WIVERN winds in WRF model: an application to the outstanding case of the Medicane Ianos” by Stefano Federico, Rosa Claudia Torcasio, Claudio Transerici, Mario Montopoli, Cinzia Cambiotti, Francesco Manconi, Alessandro Battaglia, and Maryam Pourshamsi

The paper simulates the expected improvement from assimilating line of sight Doppler winds from the future WIVERN mission for a medicane case study. Pseudo observations are produced by a selected member in an ensemble of 4 km WRF simulations downscaled from the ECMWF EPS. The cyclone track is found to improve in an ensemble with 3h or 24h data assimilation cycles compared to a control ensemble. The pressure, wind and precipitation are also impacted and the difference is reduced with respect to the selected member. The benefits of data assimilation are weakly affected by an increase in observational error estimate but clearly limited when using a generic instead of a cyclone specific background error.

The paper is generally interesting and presents new opportunities for the prediction of medicanes. However, it tends to be too specific and miss a broad general context, while lacking depth in the interpretation of results. In other words, how should the paper benefit to a broader community? Also, the text tends to be confusing by lacking consistency and repeating concepts and results. For these reasons, the paper needs major revisions before it can be considered for publication. General and specific comments are given below to help improve the paper quality.

General comments

1. From the abstract and introduction, a general scientific context is missing to motivate the specific case study using specific data in a specific model configuration: what are current limitations, what will be new or different with WIVERN, what impact is expected for a cyclone, why a medicane? The scientific context should also be discussed in the conclusions, which currently lack references to previous studies
2. There is a contradiction between the first validation of track only due to the absence of reference intensity, then validation of wind and precipitation as well but with respect to member 42 that is best in terms of track only; either include observations to assess other variables, or clarify throughout the paper what is actually achieved by data assimilation
3. The interpretation of the results is somehow blurred: how does assimilating winds impact other variables relevant to the cyclone (e.g. clouds and thermodynamics) but also the steering wind of the cyclone environment to ultimately improve tracks?
4. Repetitions in the methods and inconsistent use of definitions (acronyms and symbols) throughout the paper make the read difficult

Specific comments

In the title is it not explicit that WIVERN has not been launched yet and pseudo observations are assimilated here

l. 4–10 This sounds like an advertisement for WIVERN and does not seem too relevant here

l. 17 improves: reduces

l. 24 largely depends (but not only)

l. 28–29 This sentence is way too narrow in the broad context of data assimilation: wind at which levels? Forecasts at which scales? In which region, context, etc.? What about other observations?

l. 36 missing (

l. 37 What is Earth Explorer 11?

l. 40–42 Some comments are expected for these numbers: e.g. what can be learned from such resolution compared to previous instruments such as Aeolus described above?

l. 43–44 Why? It is the first study dedicated to this specific task but what about previous studies using WIVERN or WRF (E)DA for Mediterranean storms or elsewhere?

l. 47 Here the introduction jumps from isolated and mostly technical sentences to more structured paragraphs about Mediterranean cyclones

l. 55 missing )

l. 57 Why Ianos?

l. 65 Any insights about WRF from this paper?

l. 85 Is a cumulus parametrization activated?

l. 94 during 24 or 48h?

l. 105 what about the 3h cycle?

l. 118 it is not well emphasized that the pseudo observations are given by the best member

l. 129 how is the scan defined for a virtual case study? Why noon rather than midnight local time?

l. 131 repetition of l. 127

l. 133 CTRL using the above terminology

l. 135 why are there three segments between two dots?

l. 137 This is not obvious from Figure 2; compute the spread?

l. 138 Figure 2

l. 139 From Flaounas et al. 2023?

l. 142 Why does the use of ERA5 explain the discrepancy?

l. 143–144 This sounds like an important motivation for using the track only and should be clarified earlier

l. 154–155 is this definition (6) of D̄?

l. 161 what should be learned from Figure 3a? It is not commented apart from the two extremes

l. 166–170 largely repeats Section 2.2

l. 171 some details on the WIVERN simulator are needed to understand this result

l. 173 remove “which”

l. 185 should be “corrected observation error sigma²_cLOS”

l. 188 how is the model error computed?

l. 193 clarify why 5 km (to match the sampling of WIVERN)

l. 195 Start with Section 3.1?

l. 200 which is which island on the map?

l. 200 distance D̄?

l. 202 I do not get the point: the cyclone track is not directly related to the model winds, so why would only the cyclone intensity be the result of a propagation through model physics?

l. 204 compare panels b) WIV3h and a) CTRL?

l. 206 where is member 42 in Fig. 6b? Closer to member 42 does not necessarily means improves, as there is no reference for intensity (l. 143–144)

l. 211 see comment on l. 129

l. 215 plotting member 42 in black on Fig. 2 with the other CTRL ensemble members (instead of Figure 3b) would make it easier to compare with Figs. 5 and 7 for the other experiments

l. 215–216 syntax

l. 216 clarify Figure 2 shows CTRL (in the text and figure caption)

l. 219–221 This is hard to see without a time evolution of MSLP as in Fig. 6

l. 221–222 how changes propagate is obscure

l. 226 and WIV3h

l. 228 for WIV3h: 100 (D̄ CTRL − D̄ exp) / D̄ CTRL = 76% (not 64% as in Table 1)

l. 233 Fig 8a

l. 234 discussed below

l. 233–241 rather than discussing individual members, the member-to-member variability could be summarized by the standard deviation around the mean distance error D̄ for each ensemble in Table 1

l. 257 in the lower troposphere

l. 259 well represented compared to what?

l. 264 why discuss the zonal wind here (vertical cross section) vs. meridional wind in the other panels (horizontal cross sections)? It is very confusing and very hard to interpret

l. 265–266 how does it relate to the number of observations in Fig. 4a?

l. 271 Why show this specific forecast time? Discussing different forecast times may help better understand how the data assimilation impacts the forecast

l. 280 how do you know it is more realistic? (see comment on l. 206)

l. 283 surface winds have just been discussed

l. 285 overplotting the cyclone track would make it clearer

l. 289 clarify it is underestimated compared to member 42 (no obs here)

l. 293 “better”: as above

l. 294 with respect to

l. 296–308 in the absence of obs, discussing the wind at a specific point does not look relevant: the local “error” in intensity and direction is due to the shift in cyclone track mainly, which is largely discussed already, rather than to the simulated cyclone intensity (that is higher in CTRL)

l. 324 please stick to the terminology defined above for the trajectory errors

l. 328 “very similar”: how much is it for WIV24h?

l. 331 what should be learned from the bias and MAE shown on Fig. 14?

l. 337–341 this suggests that the NMC choice of background error matrix is not meaningful here

l. 362 Pantillon et al. (and other authors) discuss earlier initializations, while here (12 UTC on 16 September 2020) the track error is rather moderate; what would be the improvement of WIVERN data assimilation one or two days earlier?

Table 1 Err(km) should be distance D̄

Table 2 clarify it is w.r.t. member 42

Table 3 is not referred to in the text

Figure 1 The symbols cannot be read: please zoom in

Figures 2, 3b, 5, 7, 9a-c Zooming in would greatly help here as well

Figure 4b using the same notations as in the text would be helpful (see equations 7–8)

Figure 6 what is the background ensemble? = control ensemble CTRL?

Figure 8 what do diamonds and square represent?