Bathymetric Prediction Using Multisource Gravity Data Derived From a Parallel Linked BP Neural Network

Abstract

Seafloor topography, or bathymetry, has significant economic, military, and scientific research value. High-precision seafloor topographic data are helpful for studying the geographic features of the seafloor and the structure of the Earth's crust. Moreover, the seafloor requires detailed mapping to ensure fairway safety and aid in submarine navigation.Echo sounding techniques have been classically used for mapping the seafloor. From single-beam sounding in the 1950s to multibeam sounding in the 1980s , echo sounding techniques have advanced from point measurement to surface measurement. However, it is still a difficult task to obtain global seafloor topography with traditional methods within a short period (Carron et al., 2001;Sandwell & Smith, 2001). Fortunately, with the improvements in the accuracy and density of satellite altimetry data, using gravity anomalies (GAs) obtained from altimetry data to predict depth has become a feasible approach. Dixon et al. (1983) were among the first researchers to use SEASAT altimetry data to predict depth. Sandwell (1994, 1997) used GAs to construct a global seafloor model and exceedingly filled in the blank areas in nautical charts. Since then, a growing number of global seafloor models have been released, such as ETOPO1 (

Improved Bathymetry in the South China Sea from Multisource Gravity Field Elements Using Fully Connected Neural Network

“…Calculating the power spectral density (PSD) in the radial direction is a common method for spectral analysis. Many scholars have used PSD for bathymetry model assessment since a higher PSD indicates a higher topographic signal at the same wavelength [16,26]. The amplitude of the PSD in the radial direction is calculated as 10log 10 (P), where P represents the relevant power expressed in dB.…”

“…Wu et al (2022) [25] applied a back propagation neural network to recover ocean significant wave heights from S1 SAR data. [26] and Annan and Wan (2022) [23] introduced neural networks into the bathymetric prediction field for merging gravity field elements and improved the accuracy of bathymetry in the Gulf of Guinea and Mariana Trench regions. These studies show that neural networks have the potential to contribute a technical approach for bathymetry inversion by combining multi-source gravity field elements.…”