Çiçekli (2025) Performance analysis of machine learning techniques and spectral indices of determination water surface areas using Sentinel-2B satellite imagery

Identification

• Journal: Journal of Atmospheric and Solar-Terrestrial Physics
• Year: 2025
• Date: 2025-10-18
• Authors: Sevim Yasemin Çiçekli
• DOI: 10.1016/j.jastp.2025.106662

Research Groups

Department of Geological Engineering, Faculty of Engineering, Cukurova University, Adana, Türkiye

Short Summary

This study evaluates the performance of three spectral indices (NDWI, WRI, MNDWI) and two machine learning techniques (SVM, ANN) using Sentinel-2B imagery to accurately determine the water surface area of the Catalan Reservoir, finding comparable performance and high accuracy across methods.

Objective

• To determine the water surface area of the Catalan Reservoir using three spectral indices (NDWI, WRI, MNDWI) and two supervised pixel-based classification techniques (SVM, ANN) from Sentinel-2B satellite imagery.

Study Configuration

• Spatial Scale: Catalan Reservoir, located on the Seyhan River, Adana, southern Türkiye.
• Temporal Scale: December 24, 2023 (single date imagery).

Methodology and Data

• Models used: Normalized Difference Water Index (NDWI), Water Ratio Index (WRI), Modified Normalized Difference Water Index (MNDWI), Support Vector Machines (SVM), Artificial Neural Networks (ANN).
• Data sources: Sentinel-2B MSI satellite imagery.

Main Results

• The water surface area of the Catalan Reservoir was determined with high accuracy across all methods.
• NDWI yielded an area of 69.3 km² with 98.6 % accuracy.
• WRI yielded an area of 67.7 km² with 98.3 % accuracy.
• MNDWI yielded an area of 69.9 km² with 97.6 % accuracy.
• SVM classification yielded an area of 68.3 km² with 98.6 % accuracy.
• ANN classification yielded an area of 66.7 km² with 99 % accuracy.
• Spectral indices demonstrated similar performance to classification methods, suggesting their utility as practical alternatives.

Contributions

• This study is the first to use the highest-accuracy ANN pixel-based classification algorithm in conjunction with the Water Ratio Index (WRI) for water surface area determination.
• This integration provides more accurate results for remote sensing of water resources.

Funding

Not specified in the provided text.

Citation

@article{Çiçekli2025Performance,
  author = {Çiçekli, Sevim Yasemin},
  title = {Performance analysis of machine learning techniques and spectral indices of determination water surface areas using Sentinel-2B satellite imagery},
  journal = {Journal of Atmospheric and Solar-Terrestrial Physics},
  year = {2025},
  doi = {10.1016/j.jastp.2025.106662},
  url = {https://doi.org/10.1016/j.jastp.2025.106662}
}