Zhang et al. (2026) Design and trial implementation of a continental-scale, kilometre-resolution hourly precipitation analysis for Australia using satellite, radar and gauges

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-10-30
• Authors: Yuhang Zhang, Quan Jun Wang, Andrew J. Frost, Jayaram Pudashine, Blair C. Trewin, Carlos Velasco‐Forero, Chun-Hsu Su, Vincent Villani
• DOI: 10.5194/hess-30-6701-2026

Research Groups

• Department of Infrastructure Engineering, The University of Melbourne, Parkville, VIC 3010, Australia
• College of Hydrology and Water Resources, Hohai University, Nanjing 210098, China
• Bureau of Meteorology, Sydney, NSW 1300, Australia
• Bureau of Meteorology, Melbourne, VIC 3001, Australia

Short Summary

This study designs and implements BRAIN, a continental-scale, 2 km resolution hourly precipitation analysis system for Australia, integrating Himawari satellite, national radar network, and sub-daily rain gauge data using statistical interpolation. The blended analysis significantly improves precipitation estimates, capturing finer-scale rainfall structures and intensity variations compared with individual sources and existing operational products.

Objective

• To design and implement a continental-scale, kilometre-resolution hourly precipitation analysis for Australia using satellite, radar, and gauges.
• To systematically examine the role of the Himawari CRRPH satellite rainfall product as a high-resolution background field and its interaction with radar and sub-daily gauge observations within the blended analysis.
• To apply and adapt statistical interpolation theory to Australian observational and computational constraints, enabling scalable, grid-cell-wise and time-step-wise blending at kilometre and hourly resolutions with low computational overhead.

Study Configuration

• Spatial Scale: Continental-scale Australia, 2 km spatial resolution.
• Temporal Scale: Hourly temporal resolution. Trial period: 2022–2023.

Methodology and Data

• Models used: Statistical Interpolation (Optimal Interpolation, OI) framework, Kalman filtering for bias correction, Ordinary Kriging for spatial interpolation of bias fields.
• Data sources:
  • Satellite: Himawari-based Convective Rainfall Rate from Cloud Physical Properties (CRRPH v2.1) geostationary satellite rainfall estimates (2 km, 10 min).
  • Radar: Australian Bureau of Meteorology's national radar network (Rainfields system) radar rainfall estimates (1 km, 5 min).
  • Observation: Sub-daily rain gauges (15/30 min data aggregated to hourly) from the Bureau of Meteorology and other jurisdictional agencies (2155 stations for blending, ~2000 for independent validation). Independent daily rain gauge stations (3659 stations) for daily evaluation.

Main Results

• Bias correction significantly reduced systematic errors in satellite and radar precipitation estimates, making their error distributions narrower and more symmetric. Satellite data showed latitude-dependent overestimation in tropical regions, while radar exhibited localized errors.
• Error variance for both satellite and radar increased quadratically with precipitation intensity, with satellite data showing higher error variance. Satellite error correlation declined exponentially with distance, exhibiting region-specific patterns.
• The three-source blended product (BR-SRG) consistently achieved higher correlations, lower Root Mean Square Error (RMSE), and smaller biases compared with individual satellite (raw and bias-corrected), radar (operational and recalibrated), and simple gauge-based Inverse Distance Weighting (IDW) interpolation.
• BR-SRG generally outperformed the two-source satellite-gauge blend (BR-SG), demonstrating the added value of incorporating radar observations, even in areas beyond nominal radar coverage.
• Performance improvements were more pronounced under high gauge-density conditions, but BR-SRG maintained the best overall performance across both low- and high-density scenarios.
• Spatial analysis revealed high correlations and lower RMSE in coastal and well-observed areas, with higher RMSE in sparsely gauged regions.
• Case studies of extreme precipitation events (Cyclones Ellie and Jasper) demonstrated that BR-SRG effectively captured broad-scale rainfall patterns while retaining finer-scale spatial structures and intensity variations, which were less evident in the existing daily operational product (AGCD) and IDW.

Contributions

• Systematically examined the role of the Himawari CRRPH satellite rainfall product as a high-resolution background field and its interaction with radar and sub-daily gauge observations within a blended analysis for Australia.
• Presented a continental-scale trial implementation for Australia that integrates all currently available and operationally manageable satellite, radar, and gauge observations within a unified framework, extending previous multi-source efforts towards a potentially national operational product.
• Applied and adapted statistical interpolation theory to Australian observational and computational constraints, including an error-informed localization strategy and a simplified nearest-point representation for radar observations, enabling scalable, grid-cell-wise and time-step-wise blending at kilometre and hourly resolutions with low computational overhead.
• The framework allows radar information to contribute beyond hard coverage masks, providing smoother transitions across radar-coverage boundaries.

Funding

• Australian Bureau of Meteorology (Next Generation Rainfall Project)
• National Computational Infrastructure (NCI Australia)

Citation

@article{Zhang2026Design,
  author = {Zhang, Yuhang and Wang, Quan Jun and Frost, Andrew J. and Pudashine, Jayaram and Trewin, Blair C. and Velasco‐Forero, Carlos and Su, Chun-Hsu and Villani, Vincent},
  title = {Design and trial implementation of a continental-scale, kilometre-resolution hourly precipitation analysis for Australia using satellite, radar and gauges},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-6701-2026},
  url = {https://doi.org/10.5194/hess-30-6701-2026}
}