Multifrequency Backscatter Classification of Seabed Sediments Using MBES: An Integrated Approach with Ground-Truth Validation

Moonsoo Lim¹Jeongwon Kang^2*Sunhee Hwang¹Eunho Jung³Byung-Cheol Kum²Jongsin Kim¹

• ¹Marine Research Corporation, Busan, Republic of Korea
• ²Korea Institute of Ocean Science and Technology (KIOST), Busan, Republic of Korea
• ³Korea Hydrographic and Oceanographic Agency, Busan, Republic of Korea

Accurate seabed sediment classification is essential for mapping marine geological features, assessing benthic habitat, and planning coastal infrastructure. This study investigated the utility of multifrequency multibeam echosounder (MBES) backscatter data for improving seabed sediment classification compared to traditional single-frequency approaches. MBES data were acquired at three frequencies (170, 300, and 450 kHz), and post-processed to produce frequency-specific backscatter mosaics and a composite red-green-blue image. Classification was performed using unsupervised clustering methods, including K-means and isodata clustering, with input vectors composed of normalized backscatter intensities from the three frequencies. The integrated multifrequency approach successfully identified three distinct sediment classes, which were validated using grab samples analyzed for grain size, water content, total organic carbon, and slope. These classes exhibited strong correspondence with underlying geomorphological features and local hydrodynamic regimes, confirming the influence of topography and tidal currents on sediment distribution. Lower-frequency data (170 kHz) were more sensitive to subsurface variability, while higher-frequency data (450 kHz) captured surface texture differences more effectively. The combined use of all three frequencies improved classification performance, particularly in transitional sediment zones where single-frequency methods proved ambiguous. The methodology proved robust across varying water depths, sediment types, and complex seabed terrains, aligning with recent advances in MBES-based sediment mapping and supporting its general applicability for other dynamic coastal systems. These results demonstrate that the use of multifrequency MBES backscatter data enhances the resolution and reliability of sediment classification results, providing a robust framework for high-resolution seabed mapping in dynamic coastal environments.