Zhang et al. (2026) Retrieving atmospheric thermodynamic and hydrometeor profiles using a thermodynamic-constrained Kalman filter 1D-Var framework based on ground-based microwave radiometer

Identification

• Journal: Geoscientific model development
• Year: 2026
• Date: 2026-01-15
• Authors: Q. Zhang, Tianmeng Chen, Yu Wu, Bin Deng, Junjie Yan
• DOI: 10.5194/gmd-19-505-2026

Research Groups

• Key Open Laboratory of Intelligent Meteorological Observation Technology, China Meteorological Administration, Beijing, China
• Engineering Technology Research and Development Center, China Huayun Meteorological Technology Group Co., Ltd., Beijing, China
• State Key Laboratory of Severe Weather Meteorological Science and Technology & Specialized Meteorological Support Technology Research Center, Chinese Academy of Meteorological Sciences, Beijing, China
• Hainan Provincial Meteorological Observatory, Hainan, China
• Jiangxi Weather Modification Center, Nanchang, China

Short Summary

This paper introduces a novel thermodynamic-constrained Kalman filter 1D-Var (TCKF1D-Var) framework to improve the accuracy of ground-based microwave radiometer retrievals of atmospheric thermodynamic and hydrometeor profiles, especially under cloudy and precipitating conditions. The TCKF1D-Var framework significantly reduces biases in temperature and humidity, improves hydrometeor profile realism, and enhances early warning signals for heavy rainfall compared to classical 1D-Var and ERA5 reanalysis.

Objective

• To develop a thermodynamic-constrained Kalman filter variational framework (TCKF1D-Var) that enforces moist-thermodynamic consistency and integrates a diagnostic microphysics closure to retrieve more accurate temperature, humidity, and hydrometeor profiles from ground-based microwave radiometer observations, particularly under cloudy and precipitating conditions.

Study Configuration

• Spatial Scale: 44 ground-based microwave radiometer (GMWR) sites across North China, including 7 sites with collocated radiosondes.
• Temporal Scale: Continuous atmospheric profiling; validation against EarthCARE cloud liquid water content for July 2025; analysis of eight short-duration extreme precipitation events in July 2025.

Methodology and Data

• Models used:
  • TCKF1D-Var (Thermodynamic-Constrained Kalman Filter 1D-Var) framework
  • Classical 1D-Var (One-Dimensional Variational) framework
  • RTTOV-gb (Radiative Transfer for TOVS–ground-based) as the observation operator
  • WSM3 single-moment microphysics scheme for diagnostic representation of cloud liquid water and ice profiles
  • L-BFGS (Limited-memory Broyden-Fletcher-Goldfarb-Shanno) method for cost function minimization
• Data sources:
  • Ground-based microwave radiometer (GMWR) observations (seven water vapor channels: 22.240, 23.040, 23.840, 25.440, 26.240, 27.840, and 31.400 GHz; seven oxygen channels: 51.260, 52.280, 53.860, 54.940, 55.500, 56.660, and 58.000 GHz).
  • Radiosonde observations (temperature, relative humidity, pressure) from 7 co-located stations, launched twice daily (around 23:15 UTC and 11:15 UTC).
  • ERA5 reanalysis (European Centre for Medium-Range Weather Forecasts Reanalysis version 5) for a priori atmospheric profiles.
  • EarthCARE (Earth Clouds, Aerosols and Radiation Explorer) CPRCLD2A product for cloud liquid water content (CLWC) profiles.

Main Results

• TCKF1D-Var systematically reduces temperature and humidity biases relative to ERA5 and classical 1D-Var. The largest temperature bias reductions occur above 2 km, and the strongest humidity bias reductions are observed from the surface to approximately 5.5 km.
• Temperature root-mean-square errors (RMSE) from TCKF1D-Var are comparable to ERA5 and lower than 1D-Var below 8.5 km.
• Humidity RMSE is improved over 1D-Var in the near-surface layer (0–1.5 km) but shows degradation in the mid-troposphere (approximately 1.5–4.5 km) due to vertical-resolution mismatch and channel cross-talk.
• Validation against collocated EarthCARE cloud liquid water content profiles shows that TCKF1D-Var yields the lowest biases and errors and best reproduces observed distributions, particularly in the 36–136 mg m−3 range.
• Case analyses of short-duration heavy rainfall events demonstrate that TCKF1D-Var enhances precursor signals of convection (virtual potential temperature anomaly), extending the effective lead time for early warning from approximately 6–7 hours (ERA5) to approximately 7.5–8 hours, substantially outperforming 1D-Var.

Contributions

• Development of a novel TCKF1D-Var framework that integrates thermodynamic conservation constraints and a diagnostic cloud microphysics scheme into a unified variational retrieval system.
• Introduction of virtual potential temperature as a control variable, enabling joint adjustment of temperature, humidity, pressure, and hydrometeors under moist-adiabatic constraints.
• Formulation of a ratio-based cost function that is independent of prescribed climatological background and observation error covariances, allowing for better preservation of rapidly evolving atmospheric signals.
• Demonstration of improved accuracy in retrieved temperature, humidity, and hydrometeor profiles from ground-based microwave radiometers, particularly under cloudy and precipitating conditions where classical 1D-Var methods typically degrade.
• Enhancement of early warning capabilities for high-impact weather events, such as heavy rainfall, by strengthening precursory signals of convection.

Funding

• Ministry of Science and Technology of China (grant 2024YFC3013001)
• National Natural Science Foundation of China (grant 42325501)
• Heavy Rainfall Research Foundation of China (grant BYKJ2025M24)
• Basic Research Fund of CAMS (grant 2024Z003)

Citation

@article{Zhang2026Retrieving,
  author = {Zhang, Q. and Chen, Tianmeng and Guo, Jianping and Wu, Yu and Deng, Bin and Yan, Junjie},
  title = {Retrieving atmospheric thermodynamic and hydrometeor profiles using a thermodynamic-constrained Kalman filter 1D-Var framework based on ground-based microwave radiometer},
  journal = {Geoscientific model development},
  year = {2026},
  doi = {10.5194/gmd-19-505-2026},
  url = {https://doi.org/10.5194/gmd-19-505-2026}
}