<![CDATA[Graphical Abstract Highlight Research Hydroacoustic technology was used to identify seabed substrates in real-time with the Simrad EK-15 Single Beam Echosounder. Acoustic backscatter analysis classified seabed substrates into 9 sediment types, with reflection values ranging from -28.03 dB to -20.02 dB. Machine learning models (k-NN and Random Forest) achieved 98.21% and 96.43% accuracy, enabling faster sediment classification than conventional methods. Geoacoustic analysis revealed sound speed, sediment density, acoustic impedance, and reflection coefficients, defining the physical properties of the seabed. This study supports coastal engineering, marine habitat conservation, and underwater geological mapping more effectively and efficiently. AbstractHydroacoustic technology was able to quantify the seabed substrate and can be estimated accurately and near real time on the acoustic characters of each substrate. The purpose of the research was to identify the geoacoustic characteristics and spatial mapping of the seabed substrate in Lancang Island. Acoustic data was acquired using a Simrad EK-15 Single Beam Echosounder instrument operating at 200 kHz. Sediment samples were taken using an Ekman grab, which will be used to validate the acoustic data. The results of this study indicated that the acoustic backscatter values of the seabed substrate based on the surface backscattering strength value and sediment particle size at fourteen sampling stations are -28.03 decibels to -20.02 decibels divided into 9 sediment type groups, namely medium and very coarse sand mixture; medium sand; medium, fine and coarse sand mixture; medium and fine sand mixture; fine and medium sand mixture; medium and very fine sand mixture; very fine and medium sand mixture; fine and very fine sand mixture; and fine sand. The accuracy level of k-Nearest Neighbour and Random Forest computational used has very good accuracy of 98.21 % and 96.43 % and Naevi Bayes has a lower accuracy of 58.93 %. The identified geoacoustic characteristics included the mean grain size, sound speed, density, acoustic impedance, and reflection coefficient. Faster, more effective, and efficient computational processes with high accuracy make k-Nearest Neighbour and Random Forest models the best alternative to be used as geoacoustic computational models of seafloor substrates.]]>
