Kim et al. (2025) Uncertainty quantification and optimization of precipitating hydrometeor parameters for winter precipitation in a cloud microphysics scheme

Identification

• Journal: Atmospheric Research
• Year: 2025
• Date: 2025-10-11
• Authors: Ki-Byung Kim, Kyo‐Sun Sunny Lim, Junhong Lee, Kwonil Kim, Hailong Wang, Yun Qian, Jin-Ho Yoon, Yong Hee Lee, Hyoung Gwon Choi, GyuWon Lee
• DOI: 10.1016/j.atmosres.2025.108554

Research Groups

• Department of Atmospheric Sciences, Kyungpook National University, Daegu, Republic of Korea
• Center for Atmospheric REmote Sensing (CARE), Kyungpook National University, Daegu, Republic of Korea
• BK21 Weather Extremes Education & Research Team, Kyungpook National University, Daegu, Republic of Korea
• Pacific Northwest National Laboratory, Richland, WA, USA
• Department of Environment and Energy Engineering, School of Engineering, Gwangju Institute of Science and Technology, Gwangju, Republic of Korea
• Numerical Modelling Center, Korea Meteorological Administrator, Daejeon, Republic of Korea
• Department of Preventive Medicine, College of Medicine, University of Tennessee Health Science Center, Memphis, TN, USA
• UTHSC Center for Cancer Research, University of Tennessee Health Science Center, Memphis, TN, USA

Short Summary

The study quantifies uncertainties and optimizes 13 precipitating hydrometeor parameters in the WRF Double-Moment 6-class (WDM6) microphysics scheme using ICE-POP 2018 observations, achieving up to a 30.2% reduction in precipitation forecast root mean square error through Bayesian optimization.

Objective

• To quantify the uncertainties of 13 precipitating hydrometeor parameters in the Weather Research and Forecasting (WRF) Double-Moment 6-class (WDM6) microphysics scheme and optimize these parameters to improve winter precipitation forecasts over the Korean Peninsula.

Study Configuration

• Spatial Scale: Three one-way nested domains with horizontal resolutions of 9 km, 3 km, and 1 km. The outermost domain covers the entire Korean Peninsula, and the innermost domain targets the Gangwon region. Observational data were collected at the MayHills Supersite (37.6652°N, 128.6996°E) and Bokwang-ri Community Center (37.7382°N, 128.7586°E).
• Temporal Scale: Observational data from the ICE-POP 2018 field campaign (November 2017 to April 2018). Rain data accumulated over 25.7 hours, graupel over 97 minutes, and snow over 581 minutes. Model simulations for three winter precipitation cases: Cold-Low (CL) for 11 hours on 22 January 2018; Warm-Low (WL) for 22 hours from 7 to 8 March 2018; Air-Sea Interaction (AS) for 12 hours on 15 March 2018.

Methodology and Data

• Models used:
  • Weather Research and Forecasting (WRF) model version 4.4.1.
  • WRF Double-Moment 6-class (WDM6) microphysics scheme (full double-moment version).
  • Generalized Linear Model (GLM) for parameter sensitivity analysis.
  • Bayesian optimization with a Gaussian process surrogate and Expected Improvement (EI) acquisition function.
• Data sources:
  • International Collaborative Experiments for the PyeongChang 2018 Olympic and Paralympic winter games (ICE-POP 2018) field campaign observations.
  • Two-dimensional video disdrometer (2DVD) measurements.
  • Vaisala WXT520 weather sensor.
  • Multi-Angle Snowflake Camera (MASC).
  • Automatic Weather Station (AWS) observations (687-694 stations across the Korean Peninsula).
  • European Centre for Medium-Range Weather Forecasts (ECMWF) 5th reanalysis (ERA5) for initial and boundary atmospheric conditions (0.25° × 0.25° spatial resolution, 6-hourly intervals).
  • Global Ocean Forecast System version 3.1 (GOFS 3.1) for sea surface temperature (0.08° × 0.08° spatial resolution, 6-hourly intervals).

Main Results

• A comparison between the WDM6 scheme's pre-defined parameters and ICE-POP 2018 2DVD observations revealed significant deviations, particularly in the fall velocity–diameter relationship for rain, mass–diameter relationships for snow and graupel, and shape parameters for all precipitating particles.
• A Perturbed Parameter Ensemble (PPE) of 256 simulations demonstrated substantial variability in key microphysical processes (e.g., condensation of cloud water, evaporation of cloud water, accretion of cloud water by snow, accretion of cloud water by graupel) and hydrometeor mixing ratios, especially for cloud water and snow.
• Generalized Linear Model (GLM) analysis identified the mass–diameter relationship parameter for snow (cS) as a consistently highly influential parameter across all three winter precipitation cases (contributing 44.5% in CL, 12.6% in WL, and 18.9% in AS). Other influential parameters included the mass–diameter relationship parameter for snow (dS) in CL (24.0%), the fall velocity–diameter relationship parameter for snow (aS) in WL (23.8%), and the shape parameter for rain (μR) in AS (21.4%).
• Bayesian optimization successfully identified optimal parameter sets, leading to significant reductions in the root mean square error (RMSE) of simulated precipitation: 26.9% for the CL case, 30.2% for the WL case, and 15.2% for the AS case.
• The optimized parameter sets reduced the overestimation of precipitation observed in the control simulations, particularly in regions with large positive biases. For example, the RMSE for total precipitation was reduced from 0.52 × 10⁻³ m to 0.38 × 10⁻³ m in CL, from 0.86 × 10⁻³ m to 0.60 × 10⁻³ m in WL, and from 0.46 × 10⁻³ m to 0.39 × 10⁻³ m in AS.

Contributions

• This study is the first to derive hydrometeor characteristics from an intensive field campaign (ICE-POP 2018) over the Korean Peninsula and systematically compare them with the values prescribed in the WDM6 microphysics scheme.
• It provides a comprehensive quantification of uncertainties for 13 precipitating hydrometeor parameters within the WDM6 scheme, constrained by high-resolution observational data.
• The research demonstrates the effectiveness of ensemble-based uncertainty quantification combined with Bayesian optimization as a robust methodology for improving regional winter precipitation forecasts.
• It identifies specific key microphysical parameters whose perturbations significantly impact precipitation simulations, highlighting the case- and region-specific nature of these influences.
• The findings offer a pathway for improving the performance of operational weather prediction models (e.g., Korean Integrated Model, Korea Air Force-WRF) that utilize similar bulk microphysics schemes.
• The study underscores the critical importance of incorporating detailed in-situ observations for the development and refinement of microphysics parameterization schemes and suggests the potential for dynamically varying hydrometeor characteristic parameters in future models.

Funding

• Korea Meteorological Administration Research and Development Program (Grant RS-2023-00240346)
• National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIT) (RS-2025-02363044)
• Pacific Northwest National Laboratory (operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830)

Citation

@article{Kim2025Uncertainty,
  author = {Kim, Ki-Byung and Lim, Kyo‐Sun Sunny and Lee, Junhong and Kim, Kwonil and Wang, Hailong and Qian, Yun and Yoon, Jin-Ho and Lee, Yong Hee and Choi, Hyoung Gwon and Lee, GyuWon},
  title = {Uncertainty quantification and optimization of precipitating hydrometeor parameters for winter precipitation in a cloud microphysics scheme},
  journal = {Atmospheric Research},
  year = {2025},
  doi = {10.1016/j.atmosres.2025.108554},
  url = {https://doi.org/10.1016/j.atmosres.2025.108554}
}