| Abstrak/Abstract |
Understanding mangrove vegetation structures is essential for effective ecological management. High spatialresolution remote sensing provides significant opportunities for detailed mapping of these structures. This study aimed to (1) analyze correlations between remote sensing variables and field-measured vegetation characteristics, (2) develop rule sets for mangrove structure classification using geographic object-based image analysis (GEOBIA) with high-resolution imagery, and (3) evaluate classification accuracy. The analysis used Pleiades-1 satellite imagery and airborne LiDAR data from the Ratai Bay mangrove area, Lampung, Indonesia. Spectral variables, vegetation indices, texture measures, LiDAR metrics, and thematic information were integrated into the classification framework. A classification scheme based on canopy height, canopy cover, and lifeform was established. Input variables were selected using correlation-based feature selection (CFS) to identify the most relevant predictors of field-measured characteristics. Results showed that the LiDAR-derived canopy height model (CHM) and LiDAR fractional cover (LFC) were strongly correlated with actual canopy height and canopy cover, while the red band, CHM, and textural mean values from the green and red bands were highly correlated with lifeform. GEOBIA combined with a knowledge-based multilevel classification achieved an accuracy of 88–90 %, producing eight vegetation structure classes. These findings demonstrate the value of mangrove structure mapping as a scientific basis for adaptive ecosystem management, enabling managers to better understand ecosystem complexity and design targeted interventions such as multi-temporal monitoring and restoration aligned with natural zonation. |
