We thank the referee for the helpful feedback and suggestions. Below, we address each comment in detail. 

1.    Table 1: Could you use symbols such as *, # etc for the footnotes? It looks a bit confusing having the superscript numbers.

We adapted the footnote style to lower case superscript letters, following the journal guidelines.

2.    Line 63: Could you make it clear here whether you do the regridding before the front identification?

We edited the sentence to make it clear that the regridding was done before any analysis was performed: “…by following a similar approach to that described in Volosciuk et al. (2015): prior to the front detection and analysis, ERA5 is remapped to the three aforementioned coarser resolutions of 2° × 2°, 1.25° × 1.25°, and 0.9° × 0.9°.”

3.    Line 101: I’m not sure what is meant by “Using on a grid-factor”.

We removed the “on” in the sentence: “Using a grid-factor (e.g. number of high-resolution grid points per low-resolution grid point) …”. We want to convey that we cannot use a factor (e.g. 0.25° grid vs 1° gird = factor 16) to normalize the front frequency, because within the finer grid the number of frontal pixels can vary (due to frontal extend, curvature, etc.) and thus cannot be accounted for with a constant factor. 

4.    Line 103-104: I’m not sure how the resulting objects can have consistent width throughout the data. Although the radius is 300km, it depends on there being grid points within that radius in order to provide information. Could you clarify this?

That is correct. The 300 km radius used to define the frontal area masks all grid points whose centers fall within the circle. Consequently, the number of grid points within each frontal area depends on the model resolution. We acknowledge that this approach would become less reliable for very coarse data. However, our lowest-resolution model has a horizontal grid spacing of roughly 200 km, meaning that the frontal area spans on average about three grid points (≈ 600 km in width), which we consider sufficient for our analysis.

Since the radius is defined in physical space, the resulting frontal objects represent the same spatial extent across models, making their frequencies as comparable as possible. To clarify this for the reader, we have revised the sentence to: “The resulting objects have a consistent physical width of 600 km across all datasets, independent of their resolution.”

5.    Line 154: This seems inconsistent with the next paragraph that the cold fronts are more sensitive to resolution than the warm fronts.

We agree that the statement seems contradictory. The statement refers to the overall bias favoring warm fronts, which remains consistent across model resolutions. The effect described in the following paragraph—that cold front detection is more sensitive to resolution—is relatively small compared to this bias. We changed the statement to: “Comparing cold and warm fronts in more detail reveals that warm fronts are generally detected more frequently in the models. This overall bias is consistent across all model resolutions.”

6.    Line 162: Is this bias in both cold and warm fronts? Can you reference the figure panels here?

The bias is present for both cold and warm fronts alike. Previously we only mentioned the regression slopes for cold fronts in the norther regions (NEUR, NWEUR). We now added the warm front values in parentheses and added the references to the corresponding Figures: “When CORDEX is excluded, the relationship becomes stronger, with the steepness of the slope increasing from 54 and 16 (49 and 17) (Fig. 3) to 86 and 40 (76 and 36) (Fig. A2) additional fronts per 100 km for cold (warm) fronts, for NWEUR and NEUR respectively.”

7.    Figure 5: How much of the bias here is due to the precipitation values themselves, rather than the allocation to the fronts? Since you are looking at the total extreme precipitation values this could be an issue. Often analysis focuses on the proportion of extreme precipitation associated with fronts, which somewhat reduces this impact.

We agree that looking into precipitation values alone can be misleading. A paragraph on frontal extreme precipitation fractions, along with a corresponding figure in the Appendix (Fig. A6), is already included. In Fig. A6, the percentage point difference relative to ERA5 decreases for cold fronts but remains largely constant for warm fronts in CMIP6, supporting our conclusion that cold fronts benefit more from higher resolutions. We further acknowledge the impact of resolution on the total extreme precipitation in ERA5 in Fig. A3.

8.    Line 195: Are you still looking at figure 5 in this paragraph?

Yes. We added references to Fig. 5 to make this more apparent.

9.    Line 199: does this mean higher than ERA5? Can you be clear in the text what you are comparing between?

For clarity, we have added that the warm frontal precipitation over the Atlantic is compared to ERA5.

10.    Figure 7 a-p: In what way is this the “mesoscale” convergence and vorticity?

In our previous paper (Schaffer et al., 2024) we explain in detail how we split the synoptic and mesoscale dynamics. We added following sentences to the composite methodology: ”To analyze front relative circulation composites, the dynamic variable fields are split into the synoptic- and mesoscale using a spectral filter. Wavelengths longer than 1000 km make up the synoptic- and shorter wavelengths the mesoscale.” In Fig. 7a-p we display the horizontal convergence and vorticity of only the short wavelength wind field. We further added the approximate location of the frontal surfaces (based on the TFP zero-contour of the composites). We think this further increases the intuitive understanding of the cross-section figures.

11.    Figure 7: I don’t think it is discussed what the impact of the vertical resolution of the available data is on these patterns.

In theory, the vertical resolution could influence the representation of circulation patterns. Before plotting, all datasets are bilinearly interpolated to the same standard pressure levels. Since the resulting composites are already quite smooth, this interpolation has no significant impact. Even in the most extreme case—the ALADIN63 simulations, which provide only five vertical levels (925, 850, 700, 500, and 200 hPa)—the key atmospheric layers relevant for representing the frontal structure are still adequately captured. We added a sentence I the methodology section: “The extracted fields are rotated into the cross-frontal direction and bilinearly interpolated to a common set of standard pressure levels (925-200 hPa in 25 hPa steps) before computing the composites. While differences in the native vertical resolution could affect the representation of the frontal structure, the resulting composites are smooth, suggesting that interpolation and vertical resolution differences have only a minor impact.”

12.    Line 230: What figure is being referred to here?

We added references to Fig. 7 and 8: “Overall, the cold (Fig. 7) and warm frontal structure (Fig. 8) is well captured across all models …”

13.    Line 231: Is this compared to ERA5?

The biases are always evaluated by comparing the sub-ensemble mean to ERA5 with similar grid spacing.

14.    Line 232: Where is the data coming from for this -30% to 30% statement?

This value range is derived by computing the relative difference of maximum convergence of each model sub-ensemble and ERA5 with similar grid spacing. We present these percentages to give a rough estimate of how well not only the structure, but also the peak values of the circulation are represented by the models.

We hope that our revisions have improved the clarity of the manuscript and helped readers to better understand our study.

Best regards,

Armin Schaffer

We thank the referee for the helpful feedback and suggestions. Below, we address each comment in detail. 

•    Table 1: even with the footnote, I'm unclear what the two numbers of vertical levels for the CMIP6 models mean - I think that you're implying that geopotential height wasn't available for 10 of the levels the other fields were and thus only has 50 hPa spacing between 725 and 275 hPa.

This is correct. All variables except geopotential are available at 33 levels. Geopotential is missing 10 levels in the 725-275 hPa range. For clarity, we have revised the footnote to: “1000-200 (25 hPa steps), with geopotential missing 10 levels (725-275 in 50 hPa steps)”

•    Line 48: Is there some reason for this particular selection of models from the CMIP6 ensemble, particularly with the outsized influence of IPSL-CM6A-LR and IPSL-CM6A-LR-INCA on the mid-resolution sub-ensemble?

First, we analyzed all CMIP6 models for which the necessary data were available. The models were grouped into sub-ensembles based on their horizontal resolutions, which roughly cluster into three groups. The IPSL models are outliers due to their substantially different longitudinal and latitudinal grid spacing. We included these models in the mid-resolution subset because their front frequency and precipitation patterns show higher correlation with this subset than with the low-resolution sub-ensemble. This behavior is likely related to the fact that fronts (particularly warm fronts) are more sensitive to latitudinal than longitudinal resolution, as discussed in Fig. A5.

•    Line 56-57: please briefly describe the nature of the spectral filter - what is it filtering out?

We agree that the smoothing effect of the spectral filter is not clearly pointed out. We changed the sentence to: “To improve consistency, all data are spectrally filtered prior to the analysis to reduce small-scale variability.” For more details we refer the reader to our previous study (Schaffer et al., 2024), where we discuss the spectral filter in detail.

•    Line 69: is the remapping to 0.25 degree resolution for plotting/comparison a conservative remapping and not a bilinear one? A quick specification would be helpful.

We added: “… all results are remapped conservatively …”, to clarify the remapping method applied.

•    Line 89: whether or not these are "false" cold fronts is a matter of definition - I understand they aren't relevant for an examination of fronts associated with mobile systems instead of those coming from persistent land-sea contrasts though!

We fully agree with this statement. What constitutes a front depends strongly on its definition. Accordingly, we have revised the phrasing from “…the detection of false cold fronts…” to “…the detection of spurious air mass boundaries as cold fronts…”

•    Line 106: Is 6-hourly rainfall above the 99.5th percentile suitably strong enough to be considered extreme, especially when that corresponds to an expected return period of ~ 50 days?

We carefully considered the choice of threshold. What is considered “truly extreme” depends strongly on the chosen definition and the specific focus of the analysis. In the present study, we adopted the 99th percentile definition of extreme precipitation from Catto & Pfahl (2014) as a baseline and increased the threshold to represent events that are twice as rare. In our next study, we plan to classify precipitation above the 99.5th percentile as heavy and above the 99.95th percentile as extreme, corresponding to 50-day and 500-day return periods, respectively. Whether a 500-day return period can be considered “sufficiently extreme” remains subject to interpretation.

•    Line 112-114: a figure might help to illustrate what is considered a frontal object for the compositing by this definition.

There is no straightforward way to illustrate the composite method in a figure. To improve clarity, we have rewritten the description of the composite method in the manuscript, presenting it in a more intuitive, stepwise manner for the reader:

“To construct the frontal composites, we apply a multi-step procedure designed to capture the typical structures of intense frontal systems. First, we determine the precipitation associated with each frontal object by focusing on the most active 200 km segment of the front. To do so, we calculate the number of frontal grid points corresponding to a length of approximately 200 km based on the model resolution. All frontal points within a given object are then ranked by their standardized precipitation values, and the top-ranked points that together represent about 200 km of frontal length are selected. The precipitation of the frontal object is defined as the mean precipitation of these selected frontal points. This approach ensures that the composites are based on the most intense and meteorologically relevant parts of each front, independent of variations in model resolution or front length.”

•    Section 4.2: is the "frontal extreme precipitation" extreme precipitation (as in section 3.2) associated with fronts, or the precipitation associated with extreme fronts (as in section 3.3)?

Frontal extreme precipitation is defined as described in Section 3.2. At the beginning of Section 4.3, we note this explicitly for the composite analysis: “Note that these composites are based on fronts selected independently from those associated with the extreme precipitation events discussed in the previous section…” To make this clearer in Section 4.2, we have added a reference to the definition in the methodology.

We hope that our revisions have improved the clarity of the manuscript and helped readers to better understand our study.

Best regards,

Armin Schaffer

Dear Heini Wernli,

We again thank you for your constructive feedback. We address each of your comments in detail and highlight the changes to the manuscript in the attached pdf file.

We believe that our revisions enhance the manuscript’s clarity and provide readers with a clearer insight into our study.

Best regards,

Armin Schaffer