Sun et al. (2025) All-sky AMSU-A radiance data assimilation using the gain-form of Local Ensemble Transform Kalman filter within MPAS-JEDI-2.1.0: implementation, tuning, and evaluation

Identification

• Journal: Geoscientific model development
• Year: 2025
• Date: 2025-11-14
• Authors: Tao Sun, Jonathan J. Guerrette, Zhiquan Liu, Junmei Ban, Byoung‐Joo Jung, Ivette Hernández Baños, Chris Snyder
• DOI: 10.5194/gmd-18-8569-2025

Research Groups

• NSF National Center for Atmospheric Research (NCAR), Boulder, CO, USA
• Tomorrow.io, Golden, CO, USA (for one author's current affiliation)

Short Summary

This study implements and evaluates the Gain-form of Local Ensemble Transform Kalman Filter (LGETKF) within the MPAS-JEDI system for global all-sky Advanced Microwave Sounding Unit-A (AMSU-A) radiance assimilation. It demonstrates that an optimized LGETKF configuration significantly improves forecasts of moisture, wind, clouds, and precipitation, particularly in tropical regions, for up to 7 days.

Objective

• To implement and evaluate the Gain-form of Local Ensemble Transform Kalman Filter (LGETKF) within the MPAS-JEDI framework for assimilating clear-sky and all-sky microwave radiance data.
• To optimize the LGETKF's covariance inflation and localization settings for stable and robust performance in MPAS-JEDI.
• To assess the forecast impact of assimilating all-sky AMSU-A radiances over a month-long period using the optimized MPAS-JEDI LGETKF.

Study Configuration

• Spatial Scale: Global quasi-uniform grid spacing of approximately 60 km with 163,842 horizontal cells, 55 vertical levels, and a model top height of 30 km. Vertical localization scale of 6 km. Horizontal localization scale of 1200 km for non-radiance and clear-sky data, and 300 km for all-sky AMSU-A channels.
• Temporal Scale: Six-hourly cycled 80-member ensemble analysis and forecasting experiments conducted over a period of nearly one month (15 April to 14 May 2018), with forecasts extending up to 7 days.

Methodology and Data

• Models used:
  • Data Assimilation: Gain-form of Local Ensemble Transform Kalman Filter (LGETKF) within the Joint Effort for Data assimilation Integration (JEDI) framework.
  • Atmospheric Model: Model for Prediction Across Scales – Atmosphere (MPAS-A) version 7.1.
  • Radiative Transfer Model: Community Radiative Transfer Model (CRTM).
• Data sources:
  • Assimilated Observations: Radiosondes (temperature, zonal and meridional wind components, specific humidity), aircraft (temperature, zonal and meridional wind components, specific humidity), surface pressure, satellite-derived atmospheric motion vectors (AMV), Global Navigation Satellite System Radio Occultation (GNSS RO) bending angle, Advanced Microwave Sounding Unit-A (AMSU-A) radiances (clear-sky and all-sky window channels), Microwave Humidity Sounder (MHS) radiances (clear-sky).
  • Verification Data: Global Forecast System (GFS) analyses, radiosonde observations, and independent Advanced Technology Microwave Sounder (ATMS) radiances from NOAA-20.
  • Initial Ensemble Background: Time-lagged ensemble forecasts initialized from Global Ensemble Forecast System (GEFS) ensemble analyses.

Main Results

• The improved LGETKF analysis procedure significantly enhanced computational efficiency, reducing the overall time-to-completion.
• A combination of Relaxation to Prior Perturbation (RTPP, α = 0.5) and Relaxation to Prior Spread (RTPS, α = 0.9) for covariance inflation proved most effective in maintaining ensemble spread and minimizing ensemble-mean root-mean-square error (RMSE).
• A smaller horizontal localization scale of 300 km for all-sky AMSU-A radiances yielded superior short-term forecast performance compared to larger scales (600 km or 1200 km).
• Assimilating all-sky AMSU-A radiances resulted in overall positive impacts on forecasts of moisture (specific humidity) and wind components (U, V), with the largest improvements (3 %–6 % RMSE reduction for moisture) observed in tropical regions, lasting up to 7 days.
• Forecasts of clouds and precipitation also showed improvement, as indicated by better fitting to independent ATMS window channel radiances.
• Some degradation in temperature forecasts was observed, particularly in the lower troposphere around 60° S, potentially linked to model biases in simulating cold-air outbreaks.

Contributions

• First implementation and comprehensive evaluation of the LGETKF algorithm within the MPAS-JEDI framework for global all-sky radiance data assimilation.
• Demonstrated the stable and robust performance of MPAS-JEDI’s LGETKF, making it available for community research and potential operational applications.
• Provided optimized tuning configurations for covariance inflation and horizontal localization for all-sky radiance assimilation in MPAS-JEDI.
• Quantified the positive impact of all-sky AMSU-A radiance assimilation on extended-range global forecasts (up to 7 days) of key atmospheric variables (moisture, wind, clouds, precipitation).
• Identified and discussed specific challenges, such as temperature degradation in certain regions, offering insights for future model and data assimilation system improvements.

Funding

• United States Air Force (grant no. NA21OAR4310383)

Citation

@article{Sun2025Allsky,
  author = {Sun, Tao and Guerrette, Jonathan J. and Liu, Zhiquan and Ban, Junmei and Jung, Byoung‐Joo and Baños, Ivette Hernández and Snyder, Chris},
  title = {All-sky AMSU-A radiance data assimilation using the gain-form of Local Ensemble Transform Kalman filter within MPAS-JEDI-2.1.0: implementation, tuning, and evaluation},
  journal = {Geoscientific model development},
  year = {2025},
  doi = {10.5194/gmd-18-8569-2025},
  url = {https://doi.org/10.5194/gmd-18-8569-2025}
}