Su et al. (2025) Rainfall forecasting based on the multi-source data fusion and multi-dimension interpolation method for GNSS stations lacking or sparse regions

Identification

• Journal: Advances in Space Research
• Year: 2025
• Date: 2025-12-19
• Authors: Mingkun Su, J.A. Wang, Jian Wang, Cong Chen, Junna Shang, Chao Wu
• DOI: 10.1016/j.asr.2025.12.058

Research Groups

• School of Communication of Hangzhou Dianzi University
• Hubei Luojia Laboratory, Wuhan University
• GNSS Research Center, Wuhan University

Short Summary

This study proposes a novel multi-source data fusion and multi-dimension interpolation method to enable GNSS-based rainfall forecasting in regions with lacking or sparse GNSS stations, demonstrating comparable forecasting accuracy to traditional GNSS-trained models.

Objective

• To develop a new method for effective rainfall forecasting using GNSS techniques in areas where GNSS stations are lacking or sparse, overcoming the limitation of existing methods that rely heavily on historical data from dense GNSS networks.

Study Configuration

• Spatial Scale: Hong Kong, China
• Temporal Scale: 2019 to 2022 (4 years)

Methodology and Data

• Models used: Long Short-Term Memory (LSTM) model, EPGP (ERA5-PWV + GNSS-PWV + LSTM) method, MIGP (Multi-dimension interpolation + GNSS-PWV) method.
• Data sources: GNSS-PWV (Global Navigation Satellite System - Precipitable Water Vapor), ERA5-PWV (The fifth generation of European Centre for Medium-Range Weather Forecasts - Precipitable Water Vapor).

Main Results

• The proposed EPGP method achieved a True Forecasting Rate (TFR) of approximately 90.12 % and a Miss Forecast Rate (MFR) of 9.88 % for all rainfall types, comparable to results from GNSS-trained models.
• For the MIGP method, the difference between interpolated GNSS-PWV and directly observed GNSS-PWV was only 4.91 mm.
• The MIGP method demonstrated a TFR of approximately 89.83 % and a False Forecast Rate (FFR) of 20.94 %, which is equivalent to the method based on original GNSS data sets (TFR 90.32 %, FFR 20.54 %).
• Both EPGP and MIGP methods effectively resolve the challenges of applying GNSS rainfall forecasting in areas with lacking or sparse GNSS stations.

Contributions

• Proposes a novel framework (EPGP and MIGP methods) that integrates multi-source data fusion (ERA5-PWV) and multi-dimension interpolation with LSTM to extend GNSS-based rainfall forecasting capabilities to regions with limited GNSS infrastructure.
• Demonstrates that the proposed methods achieve forecasting performance comparable to traditional GNSS-trained models, addressing a critical limitation in existing GNSS rainfall forecasting techniques.

Funding

• Not specified in the provided text.

Citation

@article{Su2025Rainfall,
  author = {Su, Mingkun and Wang, J.A. and Wang, Jian and Chen, Cong and Shang, Junna and Wu, Chao},
  title = {Rainfall forecasting based on the multi-source data fusion and multi-dimension interpolation method for GNSS stations lacking or sparse regions},
  journal = {Advances in Space Research},
  year = {2025},
  doi = {10.1016/j.asr.2025.12.058},
  url = {https://doi.org/10.1016/j.asr.2025.12.058}
}