Prediction of bathymetry from satellite altimeter based gravity in the Arabian Sea: Mapping of two unnamed deep seamounts

This work attempts to predict bathymetry from satellite altimeter based gravity in the Arabian Sea. A collocated match-up database (n = 17,016) was created on Multibeam Echosounder (MBES) bathymetry and satellite gravity values (∼1 min spatial resolution) derived from remote sensing satellites. A Radial Basis Function (RBF) based Artificial Neural Network (ANN) model was developed to predict bathymetry from satellite gravity values. The ANN model was trained with variable undersea features such as seamount, knoll, abyssal plain, hill, etc. to familiarize the network with all possible geomorphic features as inputs through learning and the corresponding target outputs. The performance of the predictive model was evaluated by comparing bathymetric values with MBES datasets that were not used during the training and verification steps of the ANN model formulation. The model was then compared with MBES surveyed seamount observations (those were not used during ANN analysis) and global model bathymetry products. Results demonstrate better performance of ANN model compared to global model products for mapping of two unnamed seamounts in the Arabian Sea. These two unnamed seamounts have been predicted, mapped and their morphology is reported for the first time through this work.

► New technique for ocean bathymetry prediction from satellite altimeter based gravity. A collocated database created on MBES surveyed bathymetry and satellite gravity. ► ANN developed to predict bathymetry from gravity. Model performance assessed by comparing with MBES and global bathymetry. ► Results indicate better performance of ANN model compared to global-scale bathymetry for mapping of two unnamed seamounts. ► These unnamed seamounts have been predicted, mapped and their morphology is reported for the first time through this work.