Shirafkan et al. (2026) Improving water pixel detection in SWOT Pixel Cloud (PIXC) data over small reservoirs using machine learning methods

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-08-18
• Authors: Shayan Shirafkan, Mohammad Ali Sharifi, Mohammad J. Tourian
• DOI: 10.1016/j.jhydrol.2026.136247

Research Groups

School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Iran
Institute of Geodesy, University of Stuttgart, Germany

Short Summary

This study presents a machine learning-based framework to improve water pixel detection in SWOT Pixel Cloud (PIXC) data for reservoir monitoring applications by reducing false positive and false negative classifications.

Objective

• Investigate the feasibility of using machine learning methods to enhance water detection accuracy in SWOT PIXC data over small reservoirs.

Study Configuration

• Spatial Scale: Local scale, focusing on three reservoirs in Iran (Dez, Shahid Rajaei, and Agh Chay) with subsequent application to eight additional reservoirs.
• Temporal Scale: Long-term water occurrence prior based on Landsat 5–8 observations used as external hydrological prior.

Methodology and Data

• Models used: K-Nearest Neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) algorithms for predictive modeling and feature analysis.
• Data sources: SWOT PIXC data, Landsat 5–8 observations, in-situ measurements.

Main Results

• Machine learning models improve water detection accuracy compared to original PIXC and LakeSP products, with consistent improvements across multiple reservoirs.
• Feature analysis identifies key variables driving water/non-water separation in SWOT PIXC data, including detection uncertainty, interferometric coherence, and imaging geometry.

Contributions

• This study provides a practical framework for improving water pixel detection reliability in SWOT-derived reservoir products using machine learning methods and external hydrological priors.
• The findings highlight potential directions for refining the SWOT PIXC processing chain to reduce misclassification patterns.

Funding

• This research was funded by [project/program reference codes, e.g., "Iranian National Science Foundation (INSF)" or "University of Tehran Research Grant"].

Citation

@article{Shirafkan2026Improving,
  author = {Shirafkan, Shayan and Sharifi, Mohammad Ali and Tourian, Mohammad J.},
  title = {Improving water pixel detection in SWOT Pixel Cloud (PIXC) data over small reservoirs using machine learning methods},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.136247},
  url = {https://doi.org/10.1016/j.jhydrol.2026.136247}
}