Abstract
Mangrove ecosystems, situated in the intertidal zones of the sea, harbor a rich diversity of plant and animal species that thrive in coastal and lagoon environments. This study aims to enhance our comprehension of the intricacies within Qeshm Island's mangrove forests, located in southern Iran, through an analysis of data collected from Synthetic Aperture Radar (SAR) and optical sensors. Employing label distribution learning (LDL), a machine learning approach, this research endeavors to delineate and classify various mangrove forest types in the region. Leveraging Sentinel-1 dual-polarimetric SAR and Sentinel-2 multispectral imagery, the study evaluates six LDL algorithms, including PT-Bayes, PT-SVMs, AA-KNN, AA-BPNN, SA-IIS, and SA-BFGS, to ascertain their accuracy in classifying both pure and mixed classes. Results highlight the robust performance of LDL classification, particularly in areas exhibiting diverse species compositions, with SA-BFGS emerging as the most effective algorithm. These findings offer valuable insights into the identification of distinct mangrove communities based on their spectral and polarimetric characteristics, thereby aiding in the strategic management and conservation of these vital ecosystems.