Zhuang et al. (2025) An innovative approach to ZDR data assimilation using an ensemble Kalman filter: a proof-of-concept study

Identification

• Journal: Atmospheric Research
• Year: 2025
• Date: 2025-12-16
• Authors: Bing-Xue Zhuang, Kao‐Shen Chung, Wei‐Yu Chang, Chih-Chien Tsai, Yi‐Chiang Yu
• DOI: 10.1016/j.atmosres.2025.108703

Research Groups

• Department of Atmospheric Sciences, National Central University, Taoyuan City, Taiwan
• National Science and Technology Center for Disaster Reduction, New Taipei City, Taiwan

Short Summary

This study develops and evaluates an innovative Mean Diameter Update (MDU) approach for ZDR data assimilation using a local ensemble transform Kalman filter. By explicitly updating the mass-weighted mean diameter (Dm) of raindrops, the MDU approach leverages the strong ZDR-Dm correlation, leading to significantly improved accuracy in microphysical state analyses and short-term rainfall forecasts in both pseudo and real observation experiments.

Objective

• To develop an innovative Mean Diameter Update (MDU) approach that explicitly updates the mass-weighted mean diameter (Dm) of raindrops using assimilated differential reflectivity (ZDR) observations within a local ensemble transform Kalman filter (LETKF) radar data assimilation system.
• To evaluate the feasibility of the MDU approach and investigate its impact on analysis fields and rainfall forecasts using both pseudo and real radar observations for a Mei-Yu front case.

Study Configuration

• Spatial Scale:
  • WRF model domains: D01 (15,000 m resolution, 325 × 271 grid points), D02 (3,000 m resolution, 496 × 496 grid points), D03 (1,000 m resolution, 550 × 550 grid points).
  • Vertical levels: 52 eta levels, model top at 1,000 Pa (10 hPa).
  • Radar data resolution (degraded for assimilation): 2,000 m range and 2.0° beam for S-band radars (RCCG, RCWF); 2,250 m range and 2.25° beam for SPOL.
  • Localization radii: Horizontal 36,000 m (U, V components), 24,000 m (cloud water, cloud ice, total number concentration of cloud water, W component, perturbation geopotential, rain, snow, graupel, total number concentration of rain, mass-weighted mean diameter of rain), 12,000 m (rain, snow, graupel, total number concentration of rain, mass-weighted mean diameter of rain); Vertical 4,000 m for all variables.
• Temporal Scale:
  • Case study: Mei-Yu front event, June 6, 2022.
  • Model spin-up: 12 hours (from 1800 UTC June 5 to 0600 UTC June 6).
  • Assimilation period: 2 hours (from 0600 UTC to 0800 UTC June 6).
  • Assimilation cycling interval: 12 minutes.
  • ZDR assimilation period: Latter half of the assimilation period (from 0700 UTC to 0800 UTC).
  • Forecast period: 6 hours, initiated at 0800 UTC.

Methodology and Data

• Models used:
  • Weather Research and Forecasting (WRF) model (version 3.9.1).
  • WRF-LETKF Radar Assimilation System (WLRAS).
  • Local Ensemble Transform Kalman Filter (LETKF).
  • Microphysical parameterization scheme: WRF Double-Moment 6-category (WDM6).
  • Radiation schemes: Rapid Radiative Transfer Model (longwave), Dudhia (shortwave).
  • Planetary Boundary Layer (PBL) scheme: Yonsei University.
  • Cumulus scheme: Grell–Devenyi ensemble.
  • Observation operator: Based on Jung et al. (2008a), utilizing T-matrix scattering amplitude simulation for raindrops and Rayleigh scattering amplitude approximation for ice particles.
• Data sources:
  • Pseudo radar observations: Generated from the WRF true run output, with added zero-mean Gaussian noise (standard deviations: 3 m s⁻¹ for radial velocity (Vr), 5 dBZ for horizontal reflectivity (ZH), 0.2 dB for differential reflectivity (ZDR)).
  • Real radar observations:
    • S-band dual-polarization Doppler radars: Wufenshan (RCWF) and SPOL (National Center for Atmospheric Research).
    • S-band single-polarization radar: Chigu (RCCG).
    • Radar measurements assimilated: Radial velocity (Vr), horizontal reflectivity (ZH), differential reflectivity (ZDR).
    • Quality control (QC) applied: Terrain blocking removal, Vr unfolding, non-meteorological signal removal (using ρhv and ΦDP standard deviation), ZDR bias correction, superobbing.
  • Reanalysis data:
    • ERA5 reanalysis data: Used for initial and boundary conditions of the true run.
    • NCEP 0.25° reanalysis data: Used for initial and boundary conditions of the control run and to generate initial conditions for ensemble members.
  • Observed rainfall: Quantitative Precipitation Estimation and Segregation Using Multiple Sensors (QPESUMS) for verification.

Main Results

• Idealized Scalar Assimilation Experiment: Explicitly updating Dm with assimilated pseudo ZDR observations further improved the Dm analysis, reducing the root-mean-square errors (RMSEs) of rainwater mixing ratio (qr), total number concentration (NTr), and Dm by 40%, 18%, and 22% respectively, compared to implicit updating.
• Single-Pseudo-Observation Experiment: The MDU approach effectively leveraged the strong correlation (>0.95) between simulated ZDR and Dm, propagating more significant corrections to Dm at grid points near the assimilated observation.
• Observation System Simulation Experiments (OSSEs):
  • The MDU approach reduced analysis errors of simulated ZDR and rainwater state variables (qr, NTr) in each assimilation cycle, with ZDR analysis error improvements of 1–2% compared to non-MDU ZDR assimilation.
  • Short-term forecasts initiated with MDU-improved microphysical states showed enhanced convection, extended precipitation system duration, and more accurate rainfall accumulation patterns.
  • The MDU approach resulted in the highest probability of 6-hour rainfall exceeding 10 mm.
  • The success ratio (SR) for 2-hour rainfall accumulation was approximately 4–5% higher on average with MDU compared to non-MDU ZDR assimilation.
• Real-Observation Experiments:
  • Results were consistent with OSSEs, showing that MDU led to greater corrections in the Dm analysis and enhanced analysis error improvements for ZH and ZDR.
  • MDU further intensified 2-hour rainfall in the first forecast period and expanded the area of significant rainfall.
  • The MDU approach yielded the largest coverage of 6-hour rainfall probability greater than 20%.
  • Forecast performance (Probability of Detection and Success Ratio) with MDU was better than the control experiment throughout the forecast period and generally outperformed non-MDU ZDR assimilation in later periods.
• The MDU approach was activated at approximately 10% of the grid points where rainwater variables were updated by ZDR observations.
• The MDU approach did not lead to unreasonable values of qr and NTr.

Contributions

• Development of an innovative Mean Diameter Update (MDU) approach that explicitly leverages the direct, strong one-to-one relationship between differential reflectivity (ZDR) and the mass-weighted mean diameter (Dm) of raindrops in an ensemble Kalman filter data assimilation framework.
• Demonstrates that this explicit updating of Dm significantly enhances the accuracy of rainwater microphysical state analyses (rainwater mixing ratio, total number concentration, and Dm) and improves short-term rainfall forecasts.
• Provides a proof-of-concept through comprehensive experiments using both pseudo and real radar observations, confirming the feasibility and positive impacts of the MDU approach for convective-scale data assimilation.
• Offers a novel pathway for more accurate microphysical state estimations, which are crucial for improving quantitative precipitation forecasts.

Funding

• National Science and Technology Council of Taiwan (Research Grant 111-2111-M-008-023)
• National Center of High-Performance Computing
• National Science and Technology Center for Disaster Reduction
• Atmospheric Science Research and Application Databank

Citation

@article{Zhuang2025innovative,
  author = {Zhuang, Bing-Xue and Chung, Kao‐Shen and Chang, Wei‐Yu and Tsai, Chih-Chien and Yu, Yi‐Chiang},
  title = {An innovative approach to ZDR data assimilation using an ensemble Kalman filter: a proof-of-concept study},
  journal = {Atmospheric Research},
  year = {2025},
  doi = {10.1016/j.atmosres.2025.108703},
  url = {https://doi.org/10.1016/j.atmosres.2025.108703}
}