Rossi et al. (2025) Annual detection of wetlands using optical indices and supervised

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Springer Link (Chiba Institute of Technology)
• Year: 2025
• Date: 2025-12-12
• Authors: Hala Rossi, Ouassima El Bannoudi, Amine Bouhadi, Hakim Boulaassal, Omar El Kharki, Khadija Haboubi
• DOI: 10.1051/e3sconf/202567603001/pdf

Research Groups

[Information not available from the provided text.]

Short Summary

This study comparatively evaluates three supervised classification algorithms (Random Forest, Support Vector Machine, and Classification and Regression Tree) for wetland detection in the Tangier Tetouan Al Hoceima region using Sentinel-2 imagery and spectral indices, finding that Random Forest offers higher temporal stability and a multi-model strategy enhances detection robustness.

Objective

• To comparatively evaluate the performance of Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Tree (CART) algorithms for wetland detection in the Tangier Tetouan Al Hoceima region using Sentinel-2 imagery.

Study Configuration

• Spatial Scale: Tangier Tetouan Al Hoceima region (Mediterranean environment).
• Temporal Scale: Annual reference maps covering the 2020-2024 period.

Methodology and Data

• Models used: Random Forest (RF), Support Vector Machine (SVM), Classification and Regression Tree (CART) – all supervised classification algorithms.
• Data sources: Sentinel-2 imagery; spectral indices (Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI)).

Main Results

• The spatial extent of detected wetlands varies depending on environmental conditions and the specific classification algorithm applied.
• Random Forest (RF) demonstrates higher temporal stability in wetland detection.
• Support Vector Machine (SVM) tends to overestimate wetland coverage.
• The combined use of the three spectral indices (NDVI, NDWI, MNDWI) significantly improves overall classification accuracy.
• A multi-model strategy is suggested to enhance the robustness of wetland detection, particularly in dynamic environments facing climate change.

Contributions

• Provides a comparative evaluation of three widely used supervised classification algorithms for wetland detection in a dynamic Mediterranean environment.
• Highlights the varying performance and characteristics (e.g., temporal stability, overestimation) of different algorithms for wetland mapping.
• Demonstrates the benefit of combining multiple spectral indices for improved classification accuracy.
• Proposes a multi-model strategy as a robust approach for wetland monitoring in the context of climate change.

Funding

[Information not available from the provided text.]

Citation

@article{Rossi2025Annual,
  author = {Rossi, Hala and Bannoudi, Ouassima El and Bouhadi, Amine and Boulaassal, Hakim and Kharki, Omar El and Haboubi, Khadija},
  title = {Annual detection of wetlands using optical indices and supervised},
  journal = {Springer Link (Chiba Institute of Technology)},
  year = {2025},
  doi = {10.1051/e3sconf/202567603001/pdf},
  url = {https://doi.org/10.1051/e3sconf/202567603001/pdf}
}