Seagrass mapping using habitat suitability modeling and multibeam echosounder around Redang archipelago
Climate change and anthropogenic activities have caused the degradation of seagrass ecosystems. Hence, systematic habitat mapping and identification process are required to ensure that seagrass is protected and monitored continuously. This research aims to utilize a multibeam echosounder (MBES) syst...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/102395/1/MuhammadAbdulHakimPRAZAK2022.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Climate change and anthropogenic activities have caused the degradation of seagrass ecosystems. Hence, systematic habitat mapping and identification process are required to ensure that seagrass is protected and monitored continuously. This research aims to utilize a multibeam echosounder (MBES) system, habitat suitability modeling (HSM), and image classification to produce a seagrass seascape map at the Redang archipelago. Bathymetric map, backscatter mosaic, and their associated predictors like slope, eastness, northness, curvature, gray-level co-occurrence matrix (GLCM) texture features (homogeneity, entropy, and correlation), angular range analysis (ARA) parameters (phi and characterization) were used as the predictors. All predictors were tested for different spatial resolutions (1 and 50 m) and window sizes analysis (3 × 3, 9 × 9, and 21 × 21 pixels). For HSM, three machine learning algorithms were used: maximum entropy (MaxEnt), random forest (RF), and support vector machine (SVM). For image classification, only RF was used. Seagrass occurrence data was used to train and test the seagrass habitat suitability modeling (SHSM), while seascape feature data was used to classify and validate the seafloor classification map. The results showed that both fine and coarse spatial resolution datasets produced training models with high predictive accuracy (AUC >90%). Testing models derived from MaxEnt and RF achieved the highest predictive accuracy (AUC >90%), while the SVM models had the lowest predictive accuracy (AUC <85%). Bathymetry was found to be the most influential predictor for all models. For the coarse resolution models, backscatter predictors like ARA characterization, ARA phi, GLCM texture features, and backscatter mosaic 32-bit contributed more to produce SHSM. Different window sizes analysis and coarse spatial resolution dataset produced inconsistent habitat suitability models compared to the fine spatial resolution dataset. Overall, the MBES dataset and HSM produced a detailed seagrass habitat suitability map and provided precise information on the seagrass habitat in the Redang archipelago. The improved habitat model was proposed by integrating a seafloor classification map to associate seagrass habitat suitability index and seafloor features (i.e., seagrass on fine sand, seagrass on coarse sand, fine sand, medium sand, and coarse sand). The proposed integration method produced a detailed seascape seagrass map. The information produced from this seascape seagrass map will be useful for decision-makers like the marine park authorities to manage seagrass habitats in response to anthropogenic activities and climate change. |
|---|