ICESat-2 Satellite LiDAR Bathymetry Extraction Algorithm Based on Cubic Function Fitting Prediction Interval Along Track Segments

Abstract

The 532-nm laser pulses emitted by the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) demonstrate significant potential for shallow water depth measurement. However, due to factors such as atmospheric scattering and absorption, and solar background noise, ICESat-2 data inevitably contains many noise photons. Based on the continuously changing characteristics of the seafloor topography, this article proposes an ICESat-2 bathymetry extraction algorithm based on cubic function fitting prediction interval along track segments. The key step of this algorithm is that the underwater photons are selected using the range of the mean plus or minus five times the standard deviation of the Gaussian function fitted of the photon height histogram, after which cubic function fitting is performed. When the coefficient of determination (R²) of the fitting is larger than a threshold, the 95% prediction interval is used for denoising; if it is smaller than the threshold, denoising is performed using the Gaussian function fitting of the histogram. Five research areas are selected to conduct relevant experiments. The denoising results are evaluated by in-situ bathymetry data and the manually extracted signal photons, respectively. The results show that ICESat-2 bathymetry data and in-situ bathymetry data exhibit a highly consistent trend, with both R² greater than 0.97 and RMSE less than 0.5 m. The evaluation metrics calculated based on the manually selected signal photons all exceed 93%, with F₁ scores above 96%. Compared with adaptive elevation difference thresholding algorithm, density-based spatial clustering of applications with noise, and ordering points to identify the clustering structure, the proposed algorithm exhibits the best denoising effect, and the extracted seafloor photons show good continuity without obvious noise photons.