| 21 Jul 2026
Validation of high-resolution ICON-LES in complex terrain using observations from two HEFEX field campaigns
Abstract. High-resolution atmospheric modeling in complex mountainous terrain remains challenging because multiscaleeterogeneity, thermally driven circulation, and multiscale interactions between local and synoptic forcing. In this study, we present an extensive validation of the high-resolution ICOsahedral Nonhydrostatic (ICON) model in glacierized alpine terrain using extensive observational data from two HinterEisFerner EXperiment (HEFEX) field campaigns conducted on the Hintereisferner glacier in Austria. By integrating a dense network of automatic weather stations, instrumented towers, Doppler wind lidars, and uncrewed aerial vehicle (UAV)-based vertical soundings, we evaluate the performance of ICON large-eddy simulations (LES) at a horizontal resolution of 51 m across a range of synoptic conditions and flow regimes. The model demonstrates good performance in reproducing near-surface temperature, humidity, and wind fields, as well as the vertical structure and temporal evolution of valley-scale atmospheric flows. The remaining biases in near-surface variables can largely be attributed to shallow stable boundary layer processes that remain unresolved even at this resolution. UAV and Doppler lidar observations further show that ICON captures the vertical thermodynamic structure and dominant diurnal evolution of glacier and valley wind systems well. Remaining discrepancies primarily occur during transition periods, particularly when multiple forcing mechanisms interact, where the model shows reduced skill in accurately representing the dominant flow direction and depth. Overall, these results demonstrate that ICON-LES is capable of realistically capturing three-dimensional atmospheric dynamics in complex mountainous terrain, providing a robust foundation for future investigations of atmosphere–cryosphere coupling and glacier–climate feedback in complex mountain environments.
Review for "Validation of high-resolution ICON-LES in complex terrain using observations from two HEFEX field campaigns" by Georgi et al
Summary
-----------------
This manuscript explores the ICON model's performance in LES mode over the Hintereisferner glacier, Austria. Detailed evaluation is possible due to measurement campaign data over several years for vairous weather situations.
The scientific novelty of the results is limited, because previous studies already exist with LES over the exact same glacier, however, no study has been performed with the ICON model yet.
Still, I would like to congratulate the authors on the successful ICON simulations on dx=51m, this was surely not easy to accomplish.
However, the model output strategy is insuffient for a "proper" evaluation of an LES model, because an averaging strategy is missing; and furthermore, one instantaneous value per hour is surely not represenative for such a highly transient environment. I would like to invite the authors to check again on their model output - and, if necessary - re-run the simulations to be on the safe side for a realistic and fair comparison with the observations.
Henceforth, many of tmy concerns are of methodological nature, and the authors should also take more care in defining and explaining mountain meteorology-related terms.
I also did not comment much on the results themselves - mainly, because the output-averaging issue needs to be clarified before a in-depth analysus can be done.
Hence, this manuscript requires major revisions (and potentially new simulations) to be accepted for GMD.
Major points
------------------
1) LES averaging:
Please correct me if I missed something, but there seems to be a major flaw in the model output.
Do you really write only one file per hour for your LES? And, as a follow-up question, is this output instantaneous (i.e., non-averaged)?
If you really employ an LES at a horizontal grid spacing of 50m, your flow structure is highly turbulent and one value every hour is very likely not representative for your model.
Furthermore, if you want to grasp information on the characteristics of turbulent flow, you would have to calculate a mean and variance for turbulent variables (e.g., wind components, temperature, etc) to characterize and analyze its properties correctly.
If you did not average your turbulent flow properties from your simulation results, and only write one file very hour (even 10-minute or 5-minute values would be more beneficial), it is also no surprise that your model does not capture transient regimes, as discussed in the manuscript, because these phenomena occur on timescales of an hour or less. I am afraid that the "one file hourly" per LES simulation renders your numerical data useless.
However, there are options in ICON to write high-frequency turbulent output, this function is called "meteogram" and can be set in the model's namelist. With meteogram, you can write output for each model time step (i.e., 0.25 sec for your inntermost domain) for selected points on your model domain. A good choice would be the locations of your HEFEX stations, since you mostly compare your simulation output to point observations.
A second option would be online averagring of the 3D fields, like in Umek et al (2020) and Goger et al (2022), but this needs a large chunk of coding work, because, to my current knowledge, this feature is not implemented in ICON for turbulent quantities.
As a third option, I think ICON offers the calculation of time-averaged quantites, and I assume the authors used this function for the wind fields. However, 30 minutes might be still a too large time window to grasp information on the flow structure in the mean - as discussed abive, you still do not have the variance of the flow by this function.
2) Comparison observations -- model output.
There is another concern on the comparison between observations and simulations. I assume - at least in the case of the sonic anemmeters - the observed variables are averaged over a certain period of time. How were they processed, and with which averaging persiod and algorithm?
Furhtermore, and this is a follow-up to the model data - comparing averaged and instantaneous model output data also leads to inconsistencies int he interpretation of your results. Please clairfy.
Minor comments
------------------
line 20: "unique microclimates" - What exactly makes them unique? You might also want to define the term "microclimate".
line 21: "short distances": please provide numbers
line 22: MBL: PLease use the abbrevation "MoBL" as defined by Lehner and Rotach (2018).
line 22: "The MBL is the lowest part of the troposphere, directly influenced by mountainous
terrain, where local topography and thermal effects strongly affect atmospheric flow and turbulence." - There are two relevant points missing: the influence directly by the surface and the mentioning of timescales of a few hours or less. I would recommend that you use the direct definition from Lehner and Rotach (2018) before inventing something different.
line 25: "katabtic flows": They are also thermally-driven, right? you might mention what makes them special over glaciated surfaces compared to other anabatic/katabatic flows in complex terrain (Farina and Zardi, 2023).
line 26: "interact with larger-scale pressure gradients in ways that remain poorly understood": A flow is the results of a pressure gradient, so what do you mean with "interactions"??
line 26: "remain poorly understood". This is a very bold statement. The interaction between thermally induced flows is a major body of research in mountain meteorology and is described in numerous oublications and textbooks, e.g. by Whiteman and Durran (1993), in Whiteman (2000), and Steyn et al. (2013)
line 46: Since the manuscript is supposed to be part of the TEAMx collection, You might mention the aim and purpose of the TEAMx campaign here, trying to solve these issues in part. e.g., Rotach et al (2022)
line 50: It might be fair to mention Goger et al (2025) here as well on changing glacier surfaces.
line 55: I would not call the Inn Valley "less complex terrain".
line 55 onwards: You might state your motivation for this manuscript more clearly, by honestly mentioning that there were simulations done at this very location, but with the WRF model. You can directly start mentioning ICON's potential advantages compared to the WRF model (there are plenty), and then state that a detailed validation of ICON for this particular glacier is of general interest for the scientific communtiy. Otherwise, to be provocative, it is really not clear why we need "yet another Hintereisferner LES".
line 58: Which gap?
line 60: I think at this point the Rotach et al (2022) papr is more suitable.
line 60: Is LES not inherently "high-resolution"?
line 60: and what are high-resolution observations?
line 61: Please be aware that not all your potential readers are experts in mountain meteorology. What is the glacier-valley circulation?
line 83: This list of phenomena needs a reference ... there are plenty availabel about the HEF region.
line 83: The "previous field campaign" is called HEFEX-I.
line 93,94: At which measurement height were the quantities recorded?
line 95: Do we see on the figure which stations are ultrasonic anemomenters?
line 98: These phenomena can ba also important during accumulation, for phenomena like snow drift, see Voordendag et al (2024).
line 143: "competing forcings": What do you mean by that exactly?
line 124: Which "glacier valley exchange processes" do you wish to investigate? You often mention them in a diffuse way, but do not define them anywhere.
line 160: "nonhydrostatic dynamical core, unstructured triangular grid, and terrain-following vertical coordinates" - every state-of-the-art NWP model does have these features, this does not make ICON unique.
line 163: "explicitly" [...] "resolves a wide range of turbulent eddies": LES models are designed to resolve the largest eddies (hence the name) on their grid, but not a "wide range". What do you mean by that anyway?
line 164: "reducing the reliance on paramterizations": So which paramterizations are not neessary anymore? about which grid spacings are we talking now?
line 165: misplaced reference, Dipankar et al (1015) Did not really discuss mountainous terrain the publication on idealized simulations. Perhaps a referecne on 3D turbulence over complaex terrain might be useful, like Goger et al (2018) or Rohanizagedan et al (2015)
line 166: Every other NWP model works with limited-area grids. The relevant point is the suffiecient horizontal and vertical grid spacing. Think about the scale of the phenomena you are discussing
Table1: Please do not skip information on the intermediate grids. You submitted your manuscript to Geoscientific Model Development, your results need to be reproducible!
line 171: "resolution": I guess you mean horizontal grid spacing.
lines 170--177: Please add a plot of a map of central Europe where you show your nested domains: on the one hand, the reader knows the location of yout study, and furthermore, the domain placement is important for reproducibility.
line 173: Domains are missing in the description.
line 174: "square root of...": Add the equation instead of the description with words.
line 175: So does this mean that you use offline nesting for all your domains? Did you consider spinup issues or do you start all your simulations at the same time?
line 176: Do you mean "a suffiecient numer of grid points"?
line 178: "consistent model parameter settings": What exactly do you mean by that? Some tuning paramters or paramterizations have to be adapted for decreasing horizontal grid spacing.
line 181: "validated by Dipankar (2015)": they implemented the scheme into the ICON code - and then validated it.
lines 183,184: topography smoother: which smoother did you employ? I guess you still needed to adjust the smoother's paramters for the LES runs a bit.
line 185: Out of curiosity, why RRTM and not the more modern ecRad scheme?
line 191: "default preproseccing data sources": Which ones? Do they have names/sources? did you preprocess with extpar of icontools?
line 192: "FAO digital soil maps": reference is missing
line 193: What is the resolution of the satellite imagery?
line 195: Many previous studies with ICON LES (e.g., Goger & Dipankar 2024; Sakradzija et al. 2025, etc) use the CORINE dataset for land use. Is there a specific reason why the authors decided on the coarser, more outdated GLOBCOVER dataset?
line 203: "ICON-based atmospheric fields": this is a very vague formulation. Do these fields stem from reanalysis, forecasts, or climate simulations?
lines 203-205: Just to clarify again, are the simulations in online-or offline-nesting mode? In case of an offline-nestng approach, this has relevant implications for turbulence spinup in your inntermost domains (i.e., turbulence has to develop for each domain after intialization and is not "inherited" from the outer domain).
line 220: What are "high resolution observations"? in time and/or space?
line 241: "gradients suppress turbulent mixing": please be more precide. how does this "suppression" work?
line 245: You might mention how ICON's T2m diagnostics work. You might find them in the description of the model's surface transfer scheme, likely in Louis (1979).
line 269: Did you check the model's surface temperature?
line 271: We do not see this missing shallow yet, would this discussion make more sense in the wind speed/direction section?
line 275: Computational restraints are usually not the issue for lowest model levels. On the one hand, model stability is affected, and on the other hand, usually the model's surface layer formulation is violated.
line 280-284: Well, you could perform an evelvation correction for temperature, see Simnonet et al (2026).
line 289: What is "surface damping"?
line 292: "land surface paramterization": I guess you mean the surface transfer scheme.
line 312: You might define a bit closer the term "stable synoptic regime"
line 313: "and warm air masses over the glacier, favoring thermally stable stratification": this sounds counterintuitive, please reformulate
line 318: Now you say that the model represents the katabatic flow well, but a few sections before you mentioned that is not well-developed?
line 325: strong winds at glacier terminus: this might be also due to the change of ice to rock in general?
line 346: "persistence of thermally driven downslope flow," I would take care with the formulations here - is this now katabataic forcing or something else?
line 363: I would take care with the statement on mixing in other models - they likely treat turbulence as fully paramterized, while turbulence is partly resolved in your setup.
line 387: What is a "real model bias"? Are there imaginary ones as well?
line 398-400: An analysis of vertical profiles can give you the answer....
line 424: You have 3D model data. You can answer your questions with them.
line 438: thermal forcing is also orographzy-driven, due to differential heating. Please reformulate.
line 462. It might be worth chcking if this underestimation is systematic over the entire domain, and not only over the glacier. This might be a hint that ICON's steep orography has an impact on flow - because in other models, wind speeds over complex terrain are often overestimated because of the overly smoothed topography.
line 497: You might mention that high-resolution (meteorological) input data is also a challenge for hi-res simulations.
line 504: "ICON does not account for topograohic shading": This is incorrect, there is an implementation by MeteoSwiss on topographic shading in more recent versions based on Steger et al (2023) algorithm.
line 508: Simulating stable boundary layers is not an ICON-specific problem - there is a great overview paper by Holtslag et al (2013) and Singh et al (2025) on ICON.
line 513: "SEB" Please define.
line 517: "resolve complex alpine terrain": In this case, not the terrain itself, but the land-use spatial variability needs to be resolved.
References
-------------
- Farina, S., Zardi, D. Understanding Thermally Driven Slope Winds: Recent Advances and Open Questions. Boundary-Layer Meteorol 189, 5–52 (2023). https://doi.org/10.1007/s10546-023-00821-1
- Goger, B., Rotach, M.W., Gohm, A. et al. The Impact of Three-Dimensional Effects on the Simulation of Turbulence Kinetic Energy in a Major Alpine Valley. Boundary-Layer Meteorol 168, 1–27 (2018). https://doi.org/10.1007/s10546-018-0341-y
- Goger, B., Stiperski, I., Nicholson, L. & Sauter, T.(2022) Large-eddy simulations of the atmospheric boundary layer over an Alpine glacier: Impact of synoptic flow direction and governing processes. Q J R Meteorol Soc, 148(744), 1319–1343. Available from: https://doi.org/10.1002/qj.4263
- Goger, B. & Dipankar, A. (2024) The impact of mesh size, turbulence parameterization, and land-surface-exchange scheme on simulations of the mountain boundary layer in the hectometric range. Quarterly Journal of the Royal Meteorological Society, 150(763), 3853–3873. Available from: https://doi.org/10.1002/qj.4799
- Goger, B., Stiperski, I., Ouy, M., and Nicholson, L.: Investigating the influence of changing ice surfaces on gravity wave formation impacting glacier boundary layer flow with large-eddy simulations, Weather Clim. Dynam., 6, 345–367, https://doi.org/10.5194/wcd-6-345-2025, 2025
- Holtslag, A. A. M., and Coauthors, 2013: Stable Atmospheric Boundary Layers and Diurnal Cycles: Challenges for Weather and Climate Models. Bull. Amer. Meteor. Soc., 94, 1691–1706, https://doi.org/10.1175/BAMS-D-11-00187.1
- Lehner, Manuela, and Mathias W. Rotach. 2018. "Current Challenges in Understanding and Predicting Transport and Exchange in the Atmosphere over Mountainous Terrain" Atmosphere 9, no. 7: 276. https://doi.org/10.3390/atmos9070276
- Louis, JF. A parametric model of vertical eddy fluxes in the atmosphere. Boundary-Layer Meteorol 17, 187–202 (1979). https://doi.org/10.1007/BF00117978
- Whiteman, C. D. and Doran, J. C.: The Relationship between Overlying Synoptic-Scale Flows and Winds within a Valley, J. Appl. Meteorol., 32, 1669–1682, https://doi.org/10.1175/1520-0450(1993)032<1669:TRBOSS>2.0.CO;2, 1993
- Rohanizadegan, M., Petrone, R. M., Pomeroy, J. W., & Kosovic, B. (2025). Analysis of turbulence and turbulence kinetic energy dynamics in complex terrain. Journal of Geophysical Research: Atmospheres, 130, e2023JD040558. https://doi.org/10.1029/2023JD040558
- Rotach, M. W., and Coauthors, 2022: A Collaborative Effort to Better Understand, Measure, and Model Atmospheric Exchange Processes over Mountains. Bull. Amer. Meteor. Soc., 103, E1282–E1295, https://doi.org/10.1175/BAMS-D-21-0232.1
- Sakradzija, M., Ahlgrimm, M., Beyrich, F., Heerwaarden, C., Päschke, E., Görsdorf, U., et al. (2026) Where numerical weather prediction meets large-eddy simulations in a convective boundary layer during the FESSTVaL field experiment. Quarterly Journal of the Royal Meteorological Society, 152(774), e70037. Available from: https://doi.org/10.1002/qj.70037
- Simonet, G., Rotach, M.W. & Lehner, M. (2026) Physically consistent mesoscale model evaluation in complex terrain. Quarterly Journal of the Royal Meteorological Society, 152(775), e70063. Available from: https://doi.org/10.1002/qj.70063
- Singh, S., Schmidli, J., Bašták Ďurán, I., & Westerhuis, S. (2025). Impact of the turbulence parameterization on simulations of fog over complex terrain. Journal of Geophysical Research: Atmospheres, 130, e2024JD042610. https://doi.org/10.1029/2024JD042610
- Steger, C. R., Steger, B., and Schär, C.: HORAYZON v1.2: an efficient and flexible ray-tracing algorithm to compute horizon and sky view factor, Geosci. Model Dev., 15, 6817–6840, https://doi.org/10.5194/gmd-15-6817-2022, 2022
- Steyn, D. G., Wekker, S. F. J. D., Kossmann, M., and Martilli, A.: Boundary Layers and Air Quality in Mountainous Terrain, in: Mountain Weather Research and Forecasting, edited by: Chow, F. K., De Wekker, S. F. J., and Snyder, B. J., Springer Atmospheric Sciences, Springer Netherlands, 261–289, ISBN 978-94-007-4097-6 978-94-007-4098-3, https://doi.org/10.1007/978-94-007-4098-3_5, 2013
- Umek L, Gohm A, Haid M, Ward HC, Rotach MW. Large-eddy simulation of foehn–cold pool interactions in the Inn Valley during PIANO IOP 2. Q J R Meteorol Soc. 2021; 147: 944–982. https://doi.org/10.1002/qj.3954
Voordendag, A., Goger, B., Prinz, R., Sauter, T., Mölg, T., Saigger, M., and Kaser, G.: A novel framework to investigate wind-driven snow redistribution over an Alpine glacier: combination of high-resolution terrestrial laser scans and large-eddy simulations, The Cryosphere, 18, 849–868, https://doi.org/10.5194/tc-18-849-2024, 2024