Marine habitat mapping using multibeam echosounder survey and underwater video observations: a case study from Tioman Marine Park

In recent years, there has been an increasing trend of utilizing high-resolution multibeam echosounder (MBES) datasets and supervised classification via machine learning to create marine habitat maps. The purpose of current study was threefold: (1) to extract bathymetric and backscatter derivatives...

Full description

Saved in:
Bibliographic Details
Main Authors: Muhamad, Muhammad Abdul Hakim, Che Hasan, Rozaimi, Md. Said, Najhan, Mohd. Said, Mohd. Shahmy, Razali, Raiz
Format: Conference or Workshop Item
Language:English
Published: 2023
Subjects:
Online Access:http://eprints.utm.my/107779/1/RozaimiCheHasan2023_MarineHabitatMappingusingMultibeam.pdf
http://eprints.utm.my/107779/
http://dx.doi.org/10.1088/1755-1315/1240/1/012006
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Universiti Teknologi Malaysia
Language: English
Description
Summary:In recent years, there has been an increasing trend of utilizing high-resolution multibeam echosounder (MBES) datasets and supervised classification via machine learning to create marine habitat maps. The purpose of current study was threefold: (1) to extract bathymetric and backscatter derivatives from a multibeam dataset, (2) to measure the correlation between bathymetric and backscatter derivatives, and (3) to generate a marine habitat map using the Random Forest (RF). Tioman Marine Park (TMP), which is situated Southeast China Sea. MBES surveyed area are encompassed an area of 406 km² and served as the location for the study. Based on results and analysis, fourteen (14) derivative were derived from bathymetry map and backscatter mosaic. The second step involved integrating variables and a total of 152 of habitat ground-truth data were used, derived from underwater imageries, and sediment samples, into an RF model to generate a map of the marine habitat. Based on marine habitat map, six habitat classes including sand, rock, gravel and sand, coral rubble, coral and rock, and coral were classified. The distribution of coral habitat was found to be correlated with the depth of the bathymetry in the shallow water region. Therefore, the study has reached the conclusion that the integration between MBES derivatives, ground-truth data, and RF machine learning algorithm is an effective in classifying the distribution of marine habitats, specifically the coral habitat.