He et al. (2025) A multi-band remote sensing method for small-and-medium river discharge estimation based on fused Sentinel-2/Landsat images

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-10-15
• Authors: Liying He, Hua Chen, Nie Zhou, Ming-di Li, Chao Wang, Bin Luo, Chong‐Yu Xu
• DOI: 10.1016/j.jhydrol.2025.134424

Research Groups

• State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, China
• State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
• Changjiang Institute of Survey, Planning, Design and Research, Wuhan, China
• Department of Geosciences, University of Oslo, Oslo, Norway

Short Summary

This study developed a framework for estimating discharge in small and medium rivers using fused Sentinel-2 and Landsat 8 satellite images. The framework, which integrates a modified image fusion model and a novel multi-band remote sensing (MBRS) method, demonstrated enhanced discharge inversion capabilities, particularly with a Random Forest-based MBRS model.

Objective

• To propose and validate a framework for estimating the discharge of small and medium rivers (width < 350 m) using multi-band remote sensing data derived from fused Sentinel-2 and Landsat 8 images, aiming to address spatio-temporal discontinuities in hydrological monitoring.

Study Configuration

• Spatial Scale: Small and medium rivers (width < 350 m), applied to the Jianxi River Basin.
• Temporal Scale: Aims to generate denser, high-resolution remote sensing images to extend and interpolate runoff sequences, thereby improving temporal resolution of discharge estimates.

Methodology and Data

• Models used:
  • Modified Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) for image fusion.
  • Multi-band Remote Sensing (MBRS) method for discharge estimation, implemented with:
    • Ridge Regression (MBRS-RD) for linear relationships.
    • Random Forest (MBRS-RF) for nonlinear relationships.
  • Comparison methods: Calibration/Measurement (CM) and modified CM (CMW) methods.
• Data sources:
  • Landsat 8 satellite images.
  • Sentinel-2 satellite images.

Main Results

• The improved ESTARFM method achieved higher image fusion accuracy, with correlation coefficients (R) ranging from 0.693 to 0.958, bias from -0.0068 to 0.0040, and root mean square error (RMSE) from 0.020 to 0.163 for eight indicator layers.
• Significant correlations were observed between the constructed indicators and river discharge, with Remote Sensing Gauging Pixels (RGPs) effectively focusing on critical surface feature characteristics.
• The MBRS method demonstrated low uncertainty in discharge estimation:
  • MBRS-RD model: Nash-Sutcliffe efficiency coefficient (NSE) ranged from 0.548 to 0.839, and coefficient of determination (R²) from 0.549 to 0.839.
  • MBRS-RF model: Achieved higher performance with NSE values between 0.775 and 0.965, and R² values between 0.778 and 0.974.
• Compared to the CM (NSE = 0.499–0.861) and CMW (NSE = 0.411–0.846) methods, the MBRS approach showed enhanced discharge inversion capabilities, particularly in river reaches with minimal shoals.
• Among all models, MBRS-RF delivered the best overall performance for discharge estimation.

Contributions

• Proposes a novel framework for estimating discharge in small and medium rivers by integrating fused Sentinel-2/Landsat images and a new multi-band remote sensing (MBRS) method.
• Introduces a modified ESTARFM that generates denser, high-resolution remote sensing images with improved fusion accuracy.
• Develops the MBRS method, which utilizes eight indicators of ground features, demonstrating superior performance in discharge inversion compared to existing methods (CM, CMW).
• Offers a valuable and cost-effective supplement to traditional in-situ observation networks, particularly for addressing spatio-temporal discontinuity of flow data in remote or underdeveloped areas.

Funding

Not specified in the provided text.

Citation

@article{He2025multiband,
  author = {He, Liying and Chen, Hua and Zhou, Nie and Li, Ming-di and Wang, Chao and Luo, Bin and Xu, Chong‐Yu},
  title = {A multi-band remote sensing method for small-and-medium river discharge estimation based on fused Sentinel-2/Landsat images},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134424},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134424}
}