Tiwari et al. (2026) Enhanced precipitation estimation in a Himalayan river basin through the fusion of multi-source datasets using various machine learning techniques

Identification

• Journal: Physics and Chemistry of the Earth Parts A/B/C
• Year: 2026
• Date: 2026-03-27
• Authors: Harshita Tiwari, Aradhana Thakur, Rahul Dev Garg
• DOI: 10.1016/j.pce.2026.104418

Research Groups

Department of Civil Engineering, Indian Institute of Technology Roorkee, Haridwar, Uttarakhand, India

Short Summary

This study developed a Spatially Weighted Grid-Wise Ensemble Learning framework to enhance precipitation estimation in the Budhi Gandaki basin by fusing nine gridded precipitation products and six rain gauge observations with four machine learning algorithms. The integrated framework significantly improved precipitation estimates, achieving a correlation coefficient of 0.68 compared to 0.18-0.36 for individual products.

Objective

• To develop a scalable and adaptable Spatially Weighted Grid-Wise Ensemble Learning framework for enhanced precipitation estimation in complex terrains by fusing multi-source datasets and machine learning techniques.

Study Configuration

• Spatial Scale: Budhi Gandaki basin, central Himalaya.
• Temporal Scale: Daily.

Methodology and Data

• Models used: Random Forest, XGBoost, and two other unspecified machine learning algorithms (total of four).
• Data sources: Nine gridded precipitation products (GPPs), including APHRODITE; observations from six rain gauges.

Main Results

• Individual GPP-model pairs showed varying performance, with APHRODITE coupled with Random Forest achieving an RMSE of 5.09 mm/day at station S-1, and with XGBoost achieving an RMSE of 3.94 mm/day at station S-2, outperforming other GPPs (RMSE range 8.61-10.14 mm/day).
• Satellite, reanalysis, and ensemble-based precipitation products demonstrated superior performance with increasing elevation.
• The integrated model yielded a mean RMSE of 4.00 mm/day (ranging from 2.87 to 4.80 mm/day) with a positively skewed error distribution, effectively addressing spatial and performance gaps.
• The merged precipitation product significantly improved the correlation coefficient (CC) to 0.68, compared to a range of 0.18 to 0.36 for individual GPPs.

Contributions

• Development of a novel Spatially Weighted Grid-Wise Ensemble Learning framework for robust precipitation estimation in complex mountainous terrain.
• Demonstrated adaptability and scalability of the framework across varying elevations and data conditions, providing a significant improvement over existing single-source or uniform modeling approaches.

Funding

Not explicitly stated in the provided text.

Citation

@article{Tiwari2026Enhanced,
  author = {Tiwari, Harshita and Thakur, Aradhana and Garg, Rahul Dev},
  title = {Enhanced precipitation estimation in a Himalayan river basin through the fusion of multi-source datasets using various machine learning techniques},
  journal = {Physics and Chemistry of the Earth Parts A/B/C},
  year = {2026},
  doi = {10.1016/j.pce.2026.104418},
  url = {https://doi.org/10.1016/j.pce.2026.104418}
}