the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Aug 2026
Contrasting sensitivities of Mediterranean cyclone forecasts to model resolution and physics across cyclone types
Abstract. Hazardous Mediterranean cyclones, including tropical-like medicanes, produce some of the most damaging wind and rainfall events in southern Europe, yet their prediction remains challenging because they are often small in scale and their evolution can be strongly controlled by diabatic processes. While previous research has assessed the role of diabatic processes in individual case studies and/or using resolutions that require parametrization of convection, it remains unclear whether the sensitivity of modelled Mediterranean cyclones to explicit representation of convection is robust across cyclone types. Using a consistent multi-case framework, we identify systematic rather than case-dependent sensitivities. Simulations of seven hazardous MedCys, with a range of characteristics, are performed using four limited-area configurations of the Met Office weather forecast model downscaled from ECMWF IFS analyses. The currently operational physics packages for the global (GAL9) and limited area (RAL3) versions of the model are used with 12-km and convection-permitting, 2.2-km grid spacing, respectively; an experimental package with the scale-aware convection scheme, CoMorph, is also used at both resolutions to cleanly identify resolution dependence.
Sensitivity of dynamical variables and hazards to model configuration depends strongly on cyclone type, with medicanes exhibiting substantially greater and more robust dependence than more extratropical systems. The 2.2 km CoMorph configuration consistently over-intensifies medicanes, producing much lower minimum mean-sea-level pressures, and stronger winds and gusts than other simulations, especially for medicanes Ianos and Apollo. In contrast, the 12-km GAL9 configuration generally produces the weakest medicanes and tracks that is displaced eastward or southward; these biases are reduced using the 12-km CoMorph configuration. An in-depth analysis of medicane Ianos, which deepens by at least 20 hPa more in the 2.2-km CoMorph simulation than in the other simulations is presented. This CoMorph simulation yields a Ianos with a stronger potential vorticity tower, ascent and latent heat fluxes, and an eyewall structure resembling a tropical cyclone. These results demonstrate that the forecast skill of hazardous Mediterranean cyclones depends critically on the interaction between model resolution and physical parameterisations, implying that improvements in kilometre-scale forecasting require advances in model physics as well as increased resolution.
Competing interests: One of the co-authors is a member of the editorial board of Weather and Climate Dynamics.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(47275 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4757', Anonymous Referee #1, 11 Sep 2026
-
RC2: 'Comment on egusphere-2026-4757', Anonymous Referee #2, 26 Sep 2026
The paper discusses the effects of model resolution and physical parameterizations on simulations of intense Mediterranean cyclones, including medicanes. The manuscript is generally well written, addresses an important problem, and is well connected to previous studies. However, in its current form I have several major concerns, and I recommend a major revision.
My main concerns relate to the robustness of the results and to the extent to which the experimental design supports some of the conclusions. In particular, the manuscript frequently discusses the benefits of km-scale resolution and improvements in forecast skill. Demonstrating these claims would require either (i) experiments that isolate resolution from changes in model physics, or (ii) systematic verification against appropriate observational/reference datasets. In my view, the current analysis does not yet sufficiently achieve either. Consequently, it is often difficult to determine whether the reported differences should be interpreted as resolution effects, physics effects, resolution–physics interactions, or simply sensitivity among model configurations.
Major comments:
1) I am concerned about the robustness of some of the conclusions given the sensitivity of MedCy simulations to initial conditions. The authors themselves state that “the choice of initialisation time is critical...” and later acknowledge that ensembles with perturbed initial and/or boundary conditions would ideally be required to establish confidence in the configuration sensitivities. Nevertheless, each case/configuration is represented by only a single deterministic simulation. This makes it difficult to determine whether differences among configurations exceed the intrinsic sensitivity to initial/boundary conditions. This limitation is particularly important given the manuscript's repeated use of terms such as “robust” and “systematic.”
I appreciate that ensembles for all seven cases may be computationally prohibitive. However, an ensemble analysis for Ianos, on which a substantial part of the manuscript focuses, could provide useful information on whether the very large differences between the 2.2-km configurations exceed initial-condition uncertainty. Alternatively, the authors should more clearly qualify the robustness of their conclusions and demonstrate, as far as possible, that the configuration signal exceeds known forecast variability.
2) I am not convinced that the CPS analysis is necessary for addressing the main objectives of this manuscript. It appears to have relatively little bearing on the conclusions regarding model sensitivity and performance. Moreover, the manuscript itself demonstrates substantial sensitivity of the CPS diagnosis to the selected radius, particularly for the smaller medicanes. For Ianos, changing the low-level radius from 300 km to 160 km materially changes the diagnosed thermal structure. This raises questions about comparisons across cyclones of very different spatial scales. I suggest either removing the CPS analysis, or demonstrating that the relevant conclusions are robust to the choice of radius. An adaptive cyclone-size-dependent CPS is an interesting idea that is worth considering.
3) I think considerably more emphasis should be placed on quantitative evaluation of the different model configurations. Section 4 clearly demonstrates that the simulations are sensitive to model configuration, but it is much less clear whether these differences translate into improved forecast performance. In particular, after examining Figs. 4 and 5, I found it difficult to determine which configurations perform better, whether km-scale resolution systematically improves the simulations, and whether any improvement is consistent across cyclones and throughout their lifecycles. This is important because the first point of the Conclusions states that “the benefits of km-scale resolution for the prediction of intense MedCys are improved track, intensity and hazard predictions,” whereas I do not think this conclusion is sufficiently demonstrated by the current analysis. For example, the manuscript also states around line 382 that “the GAL9 and RAL3.3 configurations most frequently yield the largest track distance error compared to IFS-AN,” which does not suggest a straightforward improvement with increasing resolution.
I therefore suggest providing more quantitative verification of the simulations, for example through mean bias, MAE/RMSE, or similar statistics for track position, MSLP, precipitation, and other relevant variables over the cyclone lifecycles, rather than focusing primarily on selected times or the spread among configurations. Where observations are available, these should preferably be used as the benchmark. Where IFS-AN or MetUM-AN are used instead, I suggest clearly distinguishing agreement with analysis from forecast still, particularly since IFS-AN also provides the initial and lateral boundary conditions and may itself contain errors. Such an analysis would make it much clearer whether the manuscript demonstrates improved prediction at km-scale resolution or primarily demonstrates sensitivity to model configuration.
4) It’s important to make clear and separate what exactly are the resolution dependent and physics dependent effects. Comparison of GAL9 and RAL3.3 combines both changes in resolution and physics (i.e. convection scheme). How do you decouple these effects? Are there other physical schemes that change as well? (e.g. boundary layer, microphysics). Same question regarding the GAL9 and CoMorph comparison and CoMorph12 vs CoMorph2.2 comparison. The authors state (lines 100-104) that the CoMorph convection scheme is coupled with other physical schemes and has undergone significant changes (w.r.t GAL9) to make it scale-aware. Therefore, it’s hard to say which effects here are due purely to grid resolution and which are due to a specific physical scheme.
I suggest that the authors explicitly define which pairwise comparisons are intended to diagnose resolution effects, physics effects, and their interaction, and revise causal language wherever these effects cannot be uniquely separated.
5) The WISHE hypothesis is frequently referred to in the work, and is consistent with previous literature. However, recent coupled simulations (https://doi.org/10.1029/2025JD046105) of Ianos have demonstrated potentially important ocean feedbacks. So In line 94 it is mentioned that the model is only coupled to land, and in line 135 the use of OSTIA, so the lack of ocean model and the implication should at the least be addressed. This seems particularly relevant because the most extreme 2.2-km CoMorph simulation is characterized by exceptionally strong winds and surface latent heat fluxes. With prescribed SST, enhanced surface heat extraction cannot cool the underlying ocean and thereby provide a negative feedback on the surface fluxes. Could the lack of this feedback contribute to the extreme over-intensification of Ianos in the 2.2-km CoMorph simulation? Moreover, possibly the model configurations showing less WISHE-like behavior are actually more accurate when compared to ground truth.
6) One of the three primary research questions posed in the Introduction is “How does the role of diabatic processes change with km-scale resolution?” However, Sections 5.2 and 5.3 restrict the detailed structural and heat-budget analyses to the two 2.2-km simulations. This comparison can provide insight into differences between RAL3.3 and CoMorph at convection-permitting resolution, but it cannot establish how the diagnosed diabatic processes change with resolution. Since CoMorph is available at both 12 and 2.2 km, comparison of the heat-budget terms between these two simulations would seem particularly valuable and directly address the stated research question. I encourage the authors to include the 12-km CoMorph simulation in this analysis, or otherwise revise the research question and associated conclusions.
Minor comments:
- Line 6: “systematic” is a bit of a push, how systematic can the results be when only a few cases and single simulations for each are considered.
- Line 16: “is” should be “are”.
- Line 47: consider changing to “especially during rapid deepening”
- Line 58: “undergone ongoing” is redundant. Change to “...have continuously improved in terms of…” or “...undergone continuous improvements in model physics…“.
- Lines 72-74: This is somewhat misleading, as diabatic processes are not directly tuned and tested in different simulations (e.g. testing different radiation schemes or surface flux parameters). Moreover, diabatic processes is a broad term and encompasses more than what is addressed in the paper.
- Line 96: At 2.2km, I think it’s more accurate to say that the model is convective permitting and does not represent convection explicitly (e.g. https://doi.org/10.3390/meteorology1040026), as smaller scales of convection are still unresolved.
- Line 115: Why the choice of 24 hr prior to intensification? Is this the result of sensitivity tests from previous works? In the paragraph above it is stated that the later the better in terms of initialization times.
- Line 186: I wouldn’t say “mainly driven”, rephrase to convey that diabatic processes are significant/important, but not the sole main drivers.
- Eq. 3 heat budget: The diabatic term (F_T) conflates all the diabatic processes to one term. The limitation with respect to the conclusions in Sect. 5 should be discussed as the different diabatic processes are not investigated individually.
- Sect. 3.1 is long and I don’t think the in depth description of all the cyclones is essential. Perhaps adding to Table 1 additional columns (i.e. max pressure drop, max 10m wind speed, max precipitation / 3h, etc.), or listing all the important description parameters in a separate table, would be more concise and informative.
- Sect 3.2. As stated in the major comments, I don’t see how the CPS analysis supports this work, or adds insights regarding the different model performances.
- Line 394: Related to major comment, I fail to see a model that produces both skillful intensity and skillful track forecasts.
- Line 421: “The impact of model configuration at the same resolution across different cases is less clear,”. This statement is important and contradicts statements of “robust” or “systematic” finding.
- Sect. 4.2. The analyses performed for PV frequency and precipitation are done over a constant radius while the storm scales change (as the authors point out in lines 432-434). Thus I would be reluctant to reach any conclusions regarding the differences between Medicanes and Ext-MedCycs without doing some normalization by size.
- Fig. 5. is really dense. Perhaps showing something similar with differences of simulations compared with benchmarks would be clearer.
- Line 495: “throughout its lifecycle throughout its lifecycle,”
- Fig. 6: Benchmarks/ground truth is really missing.
- Same for Fig 7.
- Lines 631 - 633. It’s unclear whether the difference in the 2.2 km simulation results is because of the convective processes or because of other differences between the configurations. And even if we could narrow it down to the representation of convective processes, the important question is what exactly changes and why? This is not discussed.
- Conclusions: I disagree with point 1. Intensity has NOT been shown to improve with higher resolution.
Citation: https://doi.org/10.5194/egusphere-2026-4757-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 139 | 72 | 33 | 244 | 26 | 23 |
- HTML: 139
- PDF: 72
- XML: 33
- Total: 244
- BibTeX: 26
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Contrasting sensitivities of Mediterranean cyclone forecasts to model resolution and physics across cyclone types
Authors: Andressa Andrade Cardoso, Suzanne L. Gray, Ambrogio Volonté, Ben Harvey, and Claudio Sánchez
The purpose of the article is to explore the sensitivity of seven Mediterranean cyclones to different treatment of convection and resolution. Different cyclone types, ranging from intense extratropical cyclones to medicanes, including an hybrid case, are analyzed. In my opinion, the most relevant result is the strong sensitivity of medicane simulations to horizontal grid spacing and representation of convection. Overall, the paper is well organized and represents an interesting contribution to the rapidly growing literature of medicanes. Therefore, I recommend its publication after some relatively minor revision.
MAJOR POINTS: My main criticism is that the article is too long and the main results start very late in the paper. I suggest reducing it, for example by moving Section 3.1 to an Appendix and shortening Section 4.2, which contains a very (too?) detailed description.
MINOR POINTS: