The role of grid resolution, latent heat release, and initialisation time on the simulation of Storm &Eacute;owyn and its impacts

Hauser, Seraphine; Papritz, Lukas; Wernli, Heini

doi:10.5194/egusphere-2026-5124

Preprints

https://doi.org/10.5194/egusphere-2026-5124

Preprints

21 Sep 2026

| 21 Sep 2026

Status: this preprint is open for discussion and under review for Weather and Climate Dynamics (WCD).

The role of grid resolution, latent heat release, and initialisation time on the simulation of Storm Éowyn and its impacts

Seraphine Hauser, Lukas Papritz, and Heini Wernli

Abstract. Extratropical cyclones are primary drivers of damaging winds, heavy precipitation, and coastal flooding in the mid-latitudes. Accurately predicting their track and intensity is important for early warnings, but rapidly-deepening extratropical cyclones can pose a severe forecasting challenge due to their explosive growth, fast propagation, and extreme surface weather conditions. Capturing this rapid development in numerical weather prediction model simulations depends on grid resolution, initialisation time, as well as on the representation of small-scale diabatic processes, in particular latent heating (LH). In this study, we analyse how these factors influence the track, intensity, and surface hazards of the explosively-deepening North Atlantic Storm Éowyn, which occurred in January 2025 and caused extensive wind damage across the British Isles. Using the ICOsahedral Nonhydrostatic (ICON) model in limited-area mode, we perform sensitivity experiments with varying horizontal resolution and forecast lead time. Additionally, we run experiments with scaled LH to isolate its specific role in the development of this high-impact cyclone. The ICON simulations reveal that horizontal resolution plays a critical role for cyclone intensity, with finer grid spacings producing lower central mean sea level pressure values that align more closely with observations. The higher cyclone intensity at higher resolution is linked to a stronger vertical potential vorticity (PV) tower and translates into amplified surface winds and precipitation. With increased resolution, cloud-diabatic processes produce higher low-level PV values, which are associated with more intense surface wind speeds. When LH is fully suppressed, the resulting cyclone is significantly weaker, indicating a diabatic deepening contribution to cyclone intensity of 12 % to 53 %, depending on the forecast initialisation time. The cyclone in the experiments without LH exhibits a distinct northward shift and reduced eastward propagation, such that the cyclone misses the British Isles. The reason for this behaviour in the experiments without LH is an altered steering flow due to a weakened upper-level jet and the absence of downstream ridge building. Finally, initialisation time introduces a non-monotonic sensitivity: initialisation several days prior to cyclogenesis generally yields an underestimation of cyclone intensity and surface winds. Simulations with specific initialisation times substantially underestimate intensity across different model resolutions, highlighting a high sensitivity to early-stage initial conditions and numerical challenges in the convective 'grey zone'. Nevertheless, considering the operational ECMWF ensemble, the spread in cyclone intensity includes the observed evolution already when initialised five days before cyclogenesis. While this study examines a single high-impact cyclone, the large effects of grid resolution, initialisation time, and LH on its representation emphasizes the complex challenges in modelling explosive cyclones and highlights how sensitive these forecasts are to the specific configuration of the simulations.

Received: 26 Aug 2026 – Discussion started: 21 Sep 2026

Competing interests: At least 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.

Download & links

Preprint (PDF, 29789 KB)

Supplement (12856 KB)

Download & links

Preprint (29789 KB)
Metadata XML
Supplement (12856 KB)
BibTeX
EndNote

Seraphine Hauser, Lukas Papritz, and Heini Wernli

Status: open (until 02 Nov 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-5124', Anonymous Referee #1, 13 Oct 2026 reply
Review of: The role of grid resolution, latent heat release, and initialisation time on the simulation of Storm Éowyn and its impacts
Authors: Seraphine Hauser, Lukas Papritz, and Heini Wernli

Recommendation: Major Revisions
General Overview
In this manuscript, the authors use limited-area ICON simulations to examine the sensitivity of Storm Éowyn to horizontal grid spacing, latent-heating treatment, and forecast initialisation time. The simulations use horizontal grid spacings from 50 to 3 km and are initialized and forced at the lateral boundaries with operational IFS analyses. A second set of 13-km experiments scales selected latent-heating tendencies from zero to 125% of their control values. The authors evaluate cyclone pressure, track, low-level potential vorticity, winds, gusts, precipitation, and upper-level flow evolution. They also compare the deterministic ICON simulations with operational IFS ensemble forecasts. The manuscript addresses an interesting and important problem and is well suited to Weather and Climate Dynamics. I found the paper interesting and generally well-written; additionally the figures are clear.
My main concern is that several conclusions extend beyond what the experiments and diagnostics can support. The manuscript has the potential to advance understanding of the sensitivity of cyclone simulations to model configuration and initialization, but several claims should be narrowed or supported by additional experiments and diagnostics, as outlined in the general and specific comments below..
General Comments
The resolution experiment does not necessarily isolate the effect of horizontal grid spacing. Convection is parameterized at 50, 20, and 13 km and disabled at 7 and 3 km. Time step, numerical diffusion, effective resolution, resolved orography, and the partitioning between resolved and parameterized vertical transport also change. The manuscript nonetheless attributes most differences to horizontal resolution and later concludes that explicitly resolving convective-scale processes improves cyclone intensity. The non-monotonic changes at 13 and 7 km could arise from the switch in convection representation rather than grid spacing. The 13-km run is described as being in the convective grey zone, but 7 km is also within this transition for deep convection. A 3-km grid is convection permitting, not convection resolving. The experiment cannot readily distinguish these effects. I encourage the authors to perform matched convection-on and convection-off simulations at 7 and 13 km. A scale-aware convection sensitivity would be preferable if available. If these experiments cannot be added, the manuscript should consistently describe the results as sensitivity to model configuration rather than just resolution.

The diabatic deepening contribution is not a process budget. It is defined as a normalized difference between the minimum pressures from two nonlinear simulations, which can be at different times and locations. It is sensitive to forecast length, track divergence, land interaction, lateral-boundary history, and the pressure at initialisation. It does not partition cyclone deepening into additive dry and diabatic components. This is particularly important for the range of 12–53% discussed in the manuscript. Later initialisations have a smaller denominator because part of the cyclone deepening has already occurred. This can produce a large percentage even when the absolute control–no-LH difference is not that large. Early initialisations allow more time for nonlinear trajectory divergence. The U-shaped dependence may therefore be partly a property of the metric rather than the physical importance of latent heating changes. It would be helpful to discuss this point more and compare the results with a common verification interval to assess how important of an issue this is. Pressure differences could be evaluated at matched valid times over a common verification period. These comparisons would show how much of the U-shaped behavior arises from normalization and differing integration lengths. Unless more support is provided (which could include a budget analysis), “diabatic sensitivity” would be more accurate than “diabatic contribution.” Additionally, just to note that the text defines DDC, while the equation is labeled DDR, which briefly confused me.

The hourly samples in Figure 8 are taken from the same evolving cyclone and are correlated. Their significance cannot be assessed with a standard correlation test that treats each hour as independent. The dominant cyclone life-cycle trend may explain much of the relationship. Both low-level PV and strong-wind area respond to cyclone intensity, pressure-gradient strength, frontal contraction, storm size, stratification, translation speed, and boundary-layer fluxes. Their correlation does not show that larger low-level PV causes a larger wind area. A small increase in correlation from the 50-km to the 3-km run doesn’t establish that higher resolution improves the simulated causal relationship. Because the hourly samples are serially correlated, the correlation analysis should account for autocorrelation, for example through an effective sample size or a bootstrap approach. If the authors want to assess whether the relationship extends beyond the cyclone life-cycle trend, partial correlations controlling for intensity, storm size, and time would be useful here. Otherwise, Sections 4.3 and 5.4 should describe the result as an association rather than causations.

The no-LH simulation develops a slower and more poleward track, but the mechanism is not conclusively demonstrated. Cyclone intensity, low-level circulation, divergent outflow, jet structure, downstream ridge amplification, steering flow, and propagation all evolve together. Figures 11 and 12 document differences between the simulations, but they do not establish the mechanism described in the abstract and conclusions. The accuracy of this QG diagnostic for a strongly diabatic, rapidly deepening cyclone with substantial curvature and ageostrophic flow needs to be assessed. The manuscript does not compare the diagnosed propagation direction or speed with the actual cyclone-motion vector, as I understand. It also does not quantify the residual error. A 24-h cyclone-relative composite may make the temporal ordering of the response difficult to interpret. Please compare the observed cyclone-motion vector with the direction from the cyclone center toward the diagnosed maximum cyclonic development or propagation forcing. Please note whether the magnitude and location of the forcing are consistent with the observed propagation speed and direction through time. The statement that the no-LH cyclone lacks dynamical support in the left jet-exit region is not that convincing. Jet-quadrant conceptual models are not strictly valid in a curved and evolving jet. Divergence, vertical motion, etc. should be calculated directly.

Specific Comments
Lines 29–31: Replace “can heavily intensify” with “can substantially intensify.”

Lines 39–45: Higher resolution does not “consistently” improve all cyclone properties, and kilometer-scale simulations do not fully resolve deep convection. Qualify these statements.

Lines 50–51: Low-level PV is not produced only by latent heating. Friction and turbulent mixing should be mentioned too.

Lines 63–70: Changing forecast start time does not isolate initial-state sensitivity in a limited-area model with evolving boundary conditions.

Lines 100–104: ERA5 is used as truth for the extreme cyclone. Please provide some verification references.

Lines 136–140: The text alternates between stating that the latent-heat constant is scaled and that temperature tendencies associated with phase changes are scaled. These are not necessarily the same in terms of the numerical model modifications. Please state exactly what was changed in the source code. In particular, clarify whether alpha multiplies the latent-heat constants, the diagnosed temperature tendencies, the underlying phase-change rates, or both thermal and moisture tendencies. Also please state whether positive heating and negative cooling tendencies are both scaled and document the consequences for total-water and moist-energy conservation.

Lines 149–155: Search boxes in degrees are latitude dependent. Refer to distances in kilometres.

Lines 175–177: Support the statement about strong baroclinicity with a quantitative diagnostic, such as lower-tropospheric temperature-gradient magnitude or Eady growth rate.

Lines 210–211: Can you infer a sting jet from frontal structure alone?

Lines 234–246: The resolution interpretation is further complicated by convection treatment. You may want to note this.

Lines 265–268: If the simulated pressure is too high, this is an overestimate of minimum pressure or underestimate of cyclone intensity.

Lines 275–288: This section discusses near-surface winds but analyzes 850- and 925-hPa winds. Separate low-level dynamics from surface impacts.

Lines 327–329: Both reduced and enhanced heating shift the track northward. This non-monotonic response weakens a simple heating-based explanation.

Lines 437–444: The manuscript argues that suppression of latent heating reduces low-level PV and consequently weakens the wind field. Both changes may instead be responses to the weaker cyclone, altered pressure gradient, changed track, and modified frontal structure. A PV inversion would be needed to quantify the balanced wind attributable specifically to the low-level PV anomaly. The reference to a sting jet is also not well supported without descending trajectories and appropriate thermodynamic diagnostics. Some caveats should be included here and softening of the language.

Lines 467–483: The intensity distributions are conditional on ensemble members that satisfy the cyclone-matching criterion. Please note the matched-member sample size for the initialisation and distinguish unconditional occurrence probability from intensity conditional on a matched cyclone. The conclusion that uncertainty concerned formation rather than intensity is otherwise affected by the member-selection procedure.

Lines 529–549 and 580–592: A smaller pressure error at 3 km does not demonstrate that convection is correctly resolved, and “consistently outperform” is inconsistent with the documented non-monotonic response. Please revise the conclusions to refer to model-configuration sensitivity.

Grammar and Wording Suggestions
Line 85: Replace “a great candidate” with “a suitable case.”

Line 311: Remove “We now change subject.” The transition is somewhat abrupt.

Line 686: Remove the stray comma in “1.4, hPa.”

Figure A1 caption: Replace “number of vertical numbers” with “number of vertical levels.”

You may want to use “latent heating” rather than “latent heat release” when/if the modified tendencies include both heating and cooling associated with phase changes.

I prefer to avoid using “significant” unless a statistical test is provided.

Use “convection permitting” rather than “convection resolving” for the 3- and 7-km simulations.

Figure 3: The wind-contour description in the caption has an unmatched parenthesis.

Figure 4: The caption states that the dashed orange line represents a threshold of 11.6 m/s, although the orange variable is the 24-h pressure-deepening rate.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-5124-RC1
RC2: 'Comment on egusphere-2026-5124', Anonymous Referee #2, 19 Oct 2026 reply

Comments on “ The role of grid resolution, latent heat release, and initialisation time on the simulation of Storm Éowyn and its impacts” by Seraphine Hauser, Lukas Papritz, and Heini Wernli (egusphere-2026-5124)
Recommendation: Minor Revision
General Comments
Hauser et al. (2026) made an detailed analysis on the role of model grid-resolution, latent heat release, and initialisation time on the simulation of Storm Éowyn by using the limited-area NWP version of ICON model. The main conclusions as summarized in the abstract sound physically reasonable. Overall the manuscript is well-written and the topic is of interest and importance to the community. The three parts related to grid-resolution, latent heat release, and initialization time are basically independent with each other, and the manuscript may be overwhelmingly too long for readers to read through. Possibly it is the breadth of the scope that limits the depth of mechanical analysis and interpretation: indeed in each part some very interesting issues has been raised but they might have not been fully addressed in the manuscript. For instance, the relative contributions from the numerics (including the numerical diffusion, CFL condition, and filtering) and physical-process parameterisation to the grid-resolution-dependence of the simulations, the mechanisms responsible for northward shift of cyclone track in the simulations without latent heating, could be further explored and more insights could be obtained. Then two relatively independent manuscripts would be desirable. Despite these limitations, the manuscript as it is provides a very detailed documentation of the series of model simulations and observations, while providing some useful mechanical interpretation, and hence I recommend the manuscript be accepted after minor revisions being made.
Detailed comments and suggestions:
(1) I basically agree with the conclusions of the authors obtained from the resolution simulations, particularly about the PV tower interpretation for the high-resolution experiments, though the relative role of numerical diffusion and physical processes is not fully addressed. My suggestion is that the authors may analyze (or mention only if detailed analysis is not possible during the revision) the numerics-caused PV mixing (PV homogenization) which may lead to the weakening of PV, PV-gradient, and jets in the coarser-resolution simulations.
(2) About the simulations about the role of latent heat release, is it possible for the authors to take a look at the role of latent heat release on the strength and positioning of PV gradient (and also the orientation of upper- and mid-tropospheric jet)?
(3) Lines 348 and 355, “DDR” should be “DDC”.
(4) Figures 3 5(a,b,c), Labels or lines of latitudes and longitudes could be added.

Reply

Citation: https://doi.org/10.5194/egusphere-2026-5124-RC2
RC3: 'Comment on egusphere-2026-5124', Anonymous Referee #3, 19 Oct 2026 reply

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-5124/egusphere-2026-5124-RC3-supplement.pdf
Reply

Citation: https://doi.org/10.5194/egusphere-2026-5124-RC3

Seraphine Hauser, Lukas Papritz, and Heini Wernli

Supplement

https://doi.org/10.5194/egusphere-2026-5124-supplement

Seraphine Hauser, Lukas Papritz, and Heini Wernli

Viewed

Total article views: 644 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
391	202	51	644	53	53	48

HTML: 391
PDF: 202
XML: 51
Total: 644
Supplement: 53
BibTeX: 53
EndNote: 48

Views and downloads (calculated since 21 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	188	77	21	286
Oct 2026	203	125	30	358

Cumulative views and downloads (calculated since 21 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	188	77	21	286
Oct 2026	203	125	30	358

Viewed (geographical distribution)

Total article views: 644 (including HTML, PDF, and XML) Thereof 644 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 31 Oct 2026

Short summary

This study analyses the sensitivity of numerical simulations of Storm Éowyn in January 2025 with respect to model resolution, initialisation time, and latent heating. High resolution is crucial to capture the structure and intensity of the cyclone, and the severe wind hazards that affected parts of the British Isles. In simulations with artificially reduced latent heating, the cyclone becomes less intense, and shifts northward compared to the control simulation, missing the British Isles.


Total:	0
HTML:	0
PDF:	0
XML:	0