Please see attached file.

This paper provides a valuable 19-year, extended-winter climatology of strong winds related to cyclones in northwestern Europe. The novel aspect is the use of the RF-based approach recently published by the authors to label the near-surface winds according to distinct mesoscale features that produce the strong winds: the warm jet, cold jet, cold-front convection and cold-sector winds, along with a ‘no feature’ category. This distinction indeed shows different Earth-relative and cyclone-relative characteristics, and it opens the door to understanding the predictability of cyclone-induced winds which would likely depend on the feature.

The scope fits WCD well, the text is very well written and the figures compelling. I enjoyed reading the manuscript and I overall support its progress towards publication. There are, however, some aspects needing attention and clarification before its acceptance.

Major comments

1. Some key aspects related to the RAMEFI method need to be enhanced in the manuscript, so that Part 2 can stand alone: please explain how can station data from only 12 cyclones construct a sufficient training dataset for its applicability to (i) all cyclones in the climatology, to (ii) other, gridded data, and to (iii) location-specific features, such as land/sea differences or orographic effects.
2. It is not entirely clear how the distinction is made among the WJ/CJ/CS/CFC during the initial subjective labelling in the vicinity of the front. Since the front is a narrow and slanted surface, the location of the front line changes with height, and accordingly the attribution of the windy grid points on both its sides. It is unclear how the station-based identification of convection relates to the commonly-used front detection by the 850-hPa temperature/wind/humidity gradients and why these features are mostly absent from oceanic locations. In this context I find the comment in lines 299-300 somewhat disturbing. How should one distinguish between winds caused by frontal convection from any other isolated convective downdrafts? It would be beneficial to clarify these points, so that the high uncertainty of the CFC winds can be reduced or better understood.
3. The description of the application of the RAMEFI method to both station and model data should be enhanced. It was not clear to me from the text whether the method is carried out independently for each data set, and therefore providing independent probabilities for each, and why is it designed so? Therefore I wasn’t sure if the differences among the datasets seen in Fig. 3 stem from the fact that there are stations only over land, and they are not evenly distributed, or whether there are additional biases from the application of the method to each dataset separately?
4. The fact that NF is not randomly distributed in Figs. 6 and 7 suggests that some wind is currently associated to NF even though it should be coherently linked to the cyclone and is therefore mislabeled. Please quantify the sensitivity of the results to the labelling procedure, and estimate how much of the NF signal in Figs. 3 or 5 is found in the vicinity of cyclones.

Specific comments

Line 107: what is the timestep interval?

Subsection 2.2: Please provide more information on the model run: what is the domain? Boundary conditions? How are wind gusts and cloud cover parameterized in the model? Have they been verified against stations where these data are available?

Line 164: what is the gridding radius of the tracks to produce the “track density”? also in this line it is mentioned “of all cyclones” - it is unclear if the track density is therefore normalized, and if so, what are the units in Fig. 2?

Fig. 2: add lat/lon labels, as these are referred to in the text

Line 260-261: A bit unclear if 0-10% refer only to NF and CS? And if so, CS in line 261 should be CJ? Is the correlation computed using monthly/seasonal means?

Fig. 5 and accompanying text: why are the areas east of the Alps masked as white?

Line 356: mention how many individual cyclones are composited

Line 355-356: I cannot see the shorter duration of CFC in Figs. 6 and 7. Please clarify.

Fig. 8: are the gust data available from observations as well? Have they been verified against observations?

Line 410: Mention also the other excluded regions east of the Alps

Lines 410-412: it is unclear if the cyclone-relative windy time steps use this relative domain or a 15-degree radius as inferred from line 167. Please clarify

Technical corrections

Line 157: missing “are” before “targeted”

Line 168, 241: change “consistent to” to “consistent with”

Lines 192, 201, 209, 218: change “percentage points” to “percent” or simply “%”

Line 221: add “of” after “number”

Line 228: replace first “at” with “in”

Line 242: delete “h” from “where”

Line 273: remove “.” from “i..e”

Line 343: add “.” before “24h”

Line 391: add “relatively cooler” before “land”

Line 425: change “occurrence” to “occurs”

Line 450: replace “larger” with “longer”