Díaz-Fernández et al. (2026) Cost-benefit analysis of the nesting approach in HARMONIE-AROME for a supercell outbreak case study

Identification

• Journal: Atmospheric Research
• Year: 2026
• Date: 2026-01-14
• Authors: Javier Díaz-Fernández, Carlos Calvo‐Sancho, Morelia E. López‐Reyes, Pedro Bolgiani, J. J. González-Alemán, Àlex Morata, Daniel Santos‐Muñoz, M. L. Martín
• DOI: 10.1016/j.atmosres.2026.108774

Research Groups

• Department of Applied Mathematics, Faculty of Computer Engineering, University of Valladolid, Segovia, Spain
• State Meteorological Agency (AEMET), Madrid, Spain
• Centro de Investigaciones sobre Desertificación, Consejo Superior de Investigaciones Científicas (CIDE, CSIC-UV-GVA), Climate, Atmosphere and Ocean Laboratory (Climatoc-Lab), Moncada, Valencia, Spain
• Department of Earth Physics and Astrophysics, Faculty of Physics, Complutense University of Madrid, Madrid, Spain
• Instituto de Astronomía y Meteorología (IAM), Departamento de Física, Centro Universitario de Ciencias Exactas e Ingenierías, Universidad de Guadalajara, Guadalajara, Mexico
• Danmarks Meteorologiske Institut, Copenhagen, Denmark
• Institute of Interdisciplinary Mathematics (IMI), Complutense University of Madrid, Madrid, Spain

Short Summary

This study evaluates the cost-benefit of a two-step one-way nested HARMONIE-AROME configuration versus a single-domain configuration for simulating a severe supercell outbreak. It finds that while nesting offers minor improvements in reflectivity and temperature profiles, these are insufficient to justify its 30% higher computational cost, with the single-domain approach demonstrating superior efficiency.

Objective

• To evaluate the cost-benefit trade-off of a two-step one-way nesting approach in the HARMONIE-AROME model compared to a single-domain configuration for simulating a severe supercell outbreak within the ECMWF High-Performance Computing Facilities (HPCF) environment.

Study Configuration

• Spatial Scale: The study focuses on a supercell outbreak over eastern Iberia, characterized by complex orography. Model simulations covered this region with an outer domain of 2.5 km horizontal resolution and an inner domain of 500 m horizontal resolution for the nested approach, and a single domain of 500 m horizontal resolution for the non-nested approach. Observational radar data had a 2 km horizontal resolution, complemented by point observations from weather stations.
• Temporal Scale: The study analyzed a severe supercell outbreak on 31 July 2015. Simulations covered a 24-hour period, including a 6-hour spin-up, with hourly output frequency. Observational data for validation spanned from 10:00 to 20:00 UTC, with a specific radiosonde analysis at 12:00 UTC.

Methodology and Data

• Models used:
  • HARMONIE-AROME (cycle 46h1): A non-hydrostatic model based on spectral dynamics.
  • Two configurations were compared:
    • H500_NESTED: A two-step one-way dynamic nesting approach with a parent domain at 2.5 km horizontal resolution providing initial and boundary conditions for a nested domain at 500 m horizontal resolution.
    • H500: A single-domain configuration with a uniform 500 m horizontal resolution.
  • Initial and boundary conditions were obtained from the ECMWF Integrated Forecasting System (IFS) analysis (18 km horizontal resolution, 137 vertical model levels, updated every 6 hours).
  • Both simulations used 65 hybrid sigma-pressure vertical levels.
  • Physical parameterizations included: Morcrette shortwave radiation, Rapid Radiative Transfer Model (RRTM) for longwave radiation, ICE3 microphysics scheme (without explicit hail), EDMF scheme for shallow convection, HARATU scheme for turbulence, and SURFEXv8.1 for surface processes.
  • Diagnostic radar reflectivity was computed from hydrometeor mixing ratios of the ICE-3 scheme (rain, snow, graupel) following Caumont et al. (2010).
• Data sources:
  • Radar: OPERA data hub maximum reflectivity composites (2 km horizontal resolution, 15-minute frequency), subjected to standard quality control.
  • Surface Observations: Precipitation and 2-meter temperature observations from 236 weather stations operated by the Spanish Meteorological Agency (AEMET).
  • Radiosonde: 12:00 UTC Murcia radiosonde data (from rawinsonde.com) for analyzing pre-convective kinematic and thermodynamic parameters (e.g., CAPE, CIN, LCL, LFC, WS06, SRH03, SCP).
  • Reanalysis: ERA5 reanalysis for describing the synoptic setup (500 hPa geopotential height, 2-meter dewpoint, 10-meter wind).
  • Supercell Database: Spanish Supercell Database (Martín et al., 2021) for event selection and confirmation.

Main Results

• Computational Efficiency: The H500_NESTED simulation required 250,406 System Billing Units (SBU) and 17 hours 51 minutes of runtime on ECMWF's HPCF. In contrast, the H500 simulation required 175,106 SBU (approximately 30% less) and 14 hours 29 minutes of runtime.
• Reflectivity: Both simulations generally underestimated observed radar reflectivity (H500 bias: -5.85 dBZ; H500NESTED bias: -5.78 dBZ). H500NESTED showed a slightly better depiction of convective core structure and higher reflectivity values, but overall statistical performance (RMSE: H500 5.45 dBZ, H500_NESTED 5.48 dBZ; correlation: 0.41 for both) was comparable, with no statistically significant difference (p-value = 0.27).
• 2-meter Temperature: Both models reproduced the bimodal temperature distribution. H500_NESTED showed slightly improved statistical scores (bias: -0.36 K vs. -0.47 K for H500; RMSE: 1.02 K vs. 1.15 K; correlation: 0.98 vs. 0.97). Both configurations overestimated temperatures in the 283.15–288.15 K range and underestimated them in the 293.15–308.15 K range. No statistically significant difference was found (p-value = 0.75).
• Total Precipitation: Both configurations captured the general precipitation distribution. H500NESTED had a higher bias (0.00256 m vs. 0.00225 m for H500) but a slightly lower RMSE (0.01417 m vs. 0.01465 m) and higher correlation (0.57 vs. 0.55). While nesting improved moderate rainfall representation, it introduced larger errors for extreme amounts, with a significant overestimation in the 0.08–0.1 m range (+0.0978 m for H500NESTED vs. +0.0639 m for H500). No statistically significant difference was observed (p-value = 0.87).
• Convective Parameters (Murcia Sounding): Both simulations underestimated atmospheric instability (MUCAPE: H500NESTED 3024 J/kg, H500 3285 J/kg vs. Observed 3604 J/kg) and introduced moderate convective inhibition (MUCIN: H500NESTED -56 J/kg, H500 -124 J/kg vs. Observed 0 J/kg). H500NESTED accurately captured deep-layer bulk Wind Shear (WS06: 29 m/s vs. Observed 29 m/s), while H500 underestimated it (15 m/s). Both underestimated Storm-Relative Helicity (SRH03: H500NESTED 253 m²/s², H500 125 m²/s² vs. Observed 334 m²/s²). Supercell Convective Parameter (SCP) values were considerably lower in simulations (H500_NESTED 15, H500 6 vs. Observed 24).
• Overall Assessment: The higher-resolution nesting approach provided only limited advantages in model performance for this supercell outbreak, which were not sufficient to justify its 30% higher computational cost compared to the single-domain simulation. Marginal improvements were attributed to the parent domain already capturing main mesoscale features, potential propagation of errors, and the use of identical physical parameterizations across both configurations.

Contributions

• First cost-benefit analysis of the nesting approach in the HARMONIE-AROME model for simulating a severe supercell outbreak.
• Provides practical insights into the trade-offs between computational cost and forecast skill for high-resolution Numerical Weather Prediction (NWP) in operational and research applications.
• Evaluates the HARMONIE-AROME model's performance in representing mesoscale and convective-scale processes during a complex supercell event over eastern Iberia.
• Highlights the limitations of one-way nesting and current microphysics schemes in accurately capturing extreme precipitation and specific supercell characteristics (e.g., hail production, mesocyclone intensity).

Funding

• PID2023-146344OB-I00 (CONSCIENCE) supported by the Ministerio de Ciencia, Innovación y Universidades (MICIU), AEI/10.13039/501100011033/FEDER and UE.
• ECMWF Special Projects (SPESMART and SPESVALE).

Citation

@article{DíazFernández2026Costbenefit,
  author = {Díaz-Fernández, Javier and Calvo‐Sancho, Carlos and López‐Reyes, Morelia E. and Bolgiani, Pedro and González-Alemán, J. J. and Morata, Àlex and Santos‐Muñoz, Daniel and Martín, M. L.},
  title = {Cost-benefit analysis of the nesting approach in HARMONIE-AROME for a supercell outbreak case study},
  journal = {Atmospheric Research},
  year = {2026},
  doi = {10.1016/j.atmosres.2026.108774},
  url = {https://doi.org/10.1016/j.atmosres.2026.108774}
}