Estimation of Sound Speed Profile at Southbound of South China Sea For Bathymetric Survey Using Machine Learning

Abstract Bathymetric survey is the method done to acquire the depth of sea and understand the shape and layout of the underwater seabed by hydrographic surveyor for safety of navigation. To conduct bathymetric survey one of the main components are sound speed profile. The current practice faces many difficulties like the sensor covered by biofouling over time, sensor defects occurrence and laborious efforts. Thus, conducting post processing are the suggestion with creating a new model using Machine Learning (ML). The model is created using historical In-Situ data collected by National Hydrographic Centre as responsible agency updating nautical products for safety of navigation in Malaysian water. In this study, the area of study was done in Southbound of South China Sea and propose new method of estimating Sound Speed Profile using ML. The study describes the preprocessing steps, followed with spatial distribution of Sound Speed Profile Values using Inverse Distance Weightage (IDW). Then, based on the historical data, the data are divided between training and test dataset. Training dataset were experimented with 4 ML algorithm which is Decision Tree, Random Forest, Linear Regression and Gradient Boosting. With Random Forest showing the best result with R2 value of 0.998. After fine tune the Training Dataset, the prediction is done using test dataset which bring a final result of 0.905. Although the finding not intended to replace the real time data, the model provide hydrographic surveyors with valuable tool for conducting post-processing with more accurate estimation of sound speed profile using historical in situ data.