EUVIMEDEUVIMED entdecken
Crossref

Multispectral Discrimination of Mangroves of the Indian Sundarban Using Machine Learning Techniques

Sourav BHATTACHARYYA

Acta Montanistica Slovaca · 21.03.2026 · 1034

DOI: 10.46544/ams.v30i4.14 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Acta Montanistica Slovaca, online erschienen am 22.03.2026. Verfasst von Sourav BHATTACHARYYA. Die Arbeit nennt 27 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Mangroves in the Indian Sundarban are crucial to maintaining ecological balance and require precise monitoring to address land-cover changes and vegetation health under dynamic climatic conditions. This study utilizes LANDSAT-8/9 OLI imagery and three machine learning classifiers, viz. Maximum Likelihood (ML), Support Vector Machine (SVM), and Spectral Angle Mapper (SAM), to delineate mangroves and other LULC during the pre- and post-monsoon seasons. Pre-monsoon image analysis showed more precise spectral discrimination than post-monsoon satellite data. The post-monsoon period is characterized by increased river flow, and green vegetation and crops, resulting in identical between-class spectral signatures. In contrast, exposed mudflats, dry agricultural lands, and shallow rivers improved between-class discrimination in the pre-monsoon period. ML demonstrated better performance across both seasons, achieving high mangrove accuracy (>95%), robust Kappa values (0.96 pre-monsoon, 0.94 post-monsoon), and AUC scores (0.98 and 0.96), followed by SVM. At the same time, SAM suffered significantly during the post-monsoon period due to substantial between-class spectral similarity. Seasonal water level fluctuations, changes in settlement with vegetation, and agricultural activities influenced LULC mapping. The AUC-ROC curve further validated the classifiers' efficiency. By selecting three contrasting classifiers and explicitly accounting for the limitations of Landsat-8/9 OLI data, this study provides a rigorous benchmark for evaluating classifier robustness under seasonal variability in Sundarbans. The research revealed the superiority of the traditional maximum likelihood classifier over SVM and SAM in both seasons. Further, it provides critical insights into the seasonal dynamics of mangrove ecosystems, the need for improved LULC mapping, and the conservation and management of mangroves.

Zitieren

APA

BHATTACHARYYA, S. (2026). Multispectral Discrimination of Mangroves of the Indian Sundarban Using Machine Learning Techniques. Acta Montanistica Slovaca, 1034. https://doi.org/10.46544/ams.v30i4.14

Vancouver

BHATTACHARYYA S. Multispectral Discrimination of Mangroves of the Indian Sundarban Using Machine Learning Techniques. Acta Montanistica Slovaca. 2026:1034. doi:10.46544/ams.v30i4.14

BibTeX
@article{BHATTACHARYYA2026,
  author = {BHATTACHARYYA, Sourav},
  title = {Multispectral Discrimination of Mangroves of the Indian Sundarban Using Machine Learning Techniques},
  journal = {Acta Montanistica Slovaca},
  year = {2026},
  pages = {1034},
  doi = {10.46544/ams.v30i4.14}
}
Kostenlos starten →