Improvement of Seagrass Habitat Mapping Using Integration of Habitat Suitability Model and Seafloor Classification Map

Acoustic datasets offer considerable advantages in high-resolution seagrass habitat mapping applications. However, the nature of seafloor features for seagrass habitat on a habitat suitability model (HSM) is limited due to single–species output. Classifying the seafloor features suitable for seagrass habitat using supervised classification is propitious for explaining the potential distribution of seagrass habitat. An integrated approach is adopted between HSM and the seafloor classification map, which can simultaneously illustrate predicted seagrass habitat with suitable seafloor features. Acoustic data were acquired using a multibeam echosounder (MBES) in Redang Marine Park. Multibeam echosounder predictors, including the bathymetric map, slope, eastness, northness, and curvature, backscatter mosaic 32-bit and 8-bit, grey-level co-occurrence matrix (GLCM) texture features (homogeneity, entropy, correlation), and angular range analysis (ARA) parameters (phi and characterisation), were derived. Maximum entropy (MaxEnt) algorithm was used to produce the seagrass habitat suitability model (SHSM), while random forest (RF) algorithm was utilised to produce the seafloor classification map. As for the ground-truthing, seagrass occurrence data were used to train and test the SHSM. Improved seagrass habitat mapping information was introduced by integrating habitat suitability index, suitable seafloor features for seagrass habitat (i.e., seagrass on fine sand, seagrass on coarse sand, fine sand, and coarse sand), and suitable depth for seagrass habitat. The proposed integration approach has produced a detailed map of seascape and seagrass. The seascape seagrass map will provide decision-makers, such as the marine park authority, with helpful information for managing seagrass habitats.