Li et al. (2025) Enhancing gridded climate products with third party weather data in a rainfall study from Western Australia

Identification

• Journal: Scientific Reports
• Year: 2025
• Date: 2025-10-21
• Authors: Ming Li, Quanxi Shao
• DOI: 10.1038/s41598-025-11145-0

Research Groups

• CSIRO Data61, Australia

Short Summary

This study demonstrates the transformative potential of integrating quality-controlled third-party automatic weather station (TPAWS) data to enhance gridded climate products, specifically daily rainfall estimates in southwestern Western Australia, reducing root mean square error (RMSE) by over 15% and false no-rain rates by 30%.

Objective

• Evaluate the effectiveness of integrating quality-controlled third-party automatic weather station (TPAWS) data to improve operational gridded daily rainfall products in data-sparse regions of Australia.
• Hypothesize that, with proper quality control, erroneous TPAWS data can be treated as outliers, leaving a robust dataset to enhance rainfall estimation accuracy.

Study Configuration

• Spatial Scale: Southwestern Western Australia, generating gridded rainfall fields at 1 kilometer resolution.
• Temporal Scale: Daily rainfall estimates from 2017 to 2019.

Methodology and Data

• Models used:
  • Quality Control (QC): Automated QC procedure by Li, et al. (2023) assigning confidence scores based on comparisons with Australian Gridded Climate Data (AGCD) and Radar Rainfields.
  • Daily Rainfall Estimation: Anomaly-based approach using a two-pass Barnes interpolation.
  • Comparison Methods: Thin-plate splines (TPS), second-pass Barnes interpolation, ordinary kriging (OK), and inverse distance weighting (IDW).
• Data sources:
  • Official Weather Stations: Bureau of Meteorology (BoM) stations (79 for estimation, 25 high-quality for validation).
  • Third-Party Automatic Weather Stations (TPAWS): Department of Primary Industries and Regional Development (DPIRD) managed TPAWS (~95 stations).
  • Reference Data for QC: Australian Gridded Climate Data (AGCD) and Radar Rainfields.
  • Long-term monthly average rainfall: Obtained from AGCD.

Main Results

• Incorporating DPIRD TPAWS data reduced the Root Mean Square Error (RMSE) of daily rainfall estimates by 15–20% and Mean Absolute Error (MAE) by 18.3–21.0%.
• False no-rain rates were reduced by approximately 30% overall, with a 67% reduction for rainy days (≥ 0.2 mm/day). False rain rates decreased by 5–8%.
• Stricter quality control generally yielded the best performance but also resulted in more data removal, highlighting a trade-off between accuracy and data retention.
• Accuracy improvements were most pronounced in low rainfall categories (0.2–10 mm), followed by medium (10–30 mm) and high (> 30 mm) rainfall.
• Statistically significant accuracy gains were observed at 60% (15 of 25) of validation stations, with the largest RMSE reduction of 75% at Katanning, an inland station benefiting from dense local TPAWS coverage.
• The benefits of TPAWS data were consistent across various geostatistical interpolation methods (thin-plate splines, Barnes, ordinary kriging, inverse distance weighting).
• Higher DPIRD network density was correlated with greater accuracy gains, particularly in regions with sparse official stations.

Contributions

• Uniquely evaluates the integration of quality-controlled TPAWS data as a novel source to improve operational gridded climate products in data-sparse regions of Australia.
• Provides compelling quantitative evidence of the scientific and practical value of leveraging non-traditional datasets to address data sparsity in climate science.
• Demonstrates a scalable and cost-effective framework for integrating third-party data, applicable to improving estimates of various weather variables beyond rainfall and in other regions globally.
• Highlights the critical role of quality control in effectively utilizing TPAWS data and managing the balance between data accuracy and retention.

Funding

• CAS-CSIRO Partnership program

Citation

@article{Li2025Enhancing,
  author = {Li, Ming and Shao, Quanxi},
  title = {Enhancing gridded climate products with third party weather data in a rainfall study from Western Australia},
  journal = {Scientific Reports},
  year = {2025},
  doi = {10.1038/s41598-025-11145-0},
  url = {https://doi.org/10.1038/s41598-025-11145-0}
}