Development of a Statistical Outlier Removal Method through Valid Signal Range Refinement of Airborne Bathymetric LiDAR Waveforms

This study aims to improve point cloud quality and the reliability of terrain interpretation by systematically suppressing outliers caused by waveform overfitting during airborne bathymetric LiDAR full waveform decomposition. PGD (Progressive Gaussian Decomposition) is applied to separate surface and seafloor returns and to extract individual parameters such as the centroid and width for each waveform component. The spatial similarity among neighboring waveforms is then exploited to define statistical thresholds for outlier removal and to detect anomalous valid signal ranges, followed by refinement of these ranges, which suppresses unnecessary decomposition in intervals without signal. Experimental results show that mitigating overfitting reduces the number of noise points by approximately 73% compared with the conventional processing, while preserving the spatial continuity of bottom point distribution and of surface and seafloor cross sections. These findings demonstrate that valid signal range refinement and outlier removal based on neighboring waveform statistics constitute an effective strategy for airborne bathymetric LiDAR data processing and terrain interpretation.