Sound Velocity Predictive Model Based on Physical Properties

摘要

The correlation between sediment sound velocity ( V ) and physical properties has been studied for 60 years using empirical equations, and it has been found difficult to predict V accurately. Random Forest (RF) is a scientific discipline and a method of data analysis that automates analytical model building. Here we present the implementation of RF algorithm in V prediction and sediment classification. The databases were from previously collected data in the northern South China Sea. The goal of this study is to establish a predictive model based on RF using multiple physical properties (mean grain size, porosity, wet bulk density, and water content). Compared to empirical equations, the average error of RF velocity is only 0.95%, ranging from 0.03 to 2.73%, indicating that the RF algorithm has improved the accuracy of V prediction. We also used mean decrease impurity importance to evaluate the importance of a variable and found that the most important feature in the predictive model is mean grain size. We also used the RF as a potentially useful tool for sediment classification. The classification model has up to 75% accuracy in the dataset. Multiple features, such as physical properties, sedimentary environment, and sediment source, affect the geoacoustic properties of sediments. The next goal is to use multiple features to improve the model and further improve the accuracy of sound velocity prediction and sediment classification. Plain Language Summary The sound velocity is usually predicted by empirical equations; however, the accuracy is not good, because a single‐parameter equation or a two‐parameter equation cannot fully represent the nature of sediment. We used a Random Forest algorithm combined with four physical properties (mean grain size, porosity, wet bulk density, and water content) to predict sound velocity, and the results are better than empirical equations. The results we present here indicates that a machine learning method such as Random Forest or other algorithms can be used in geoscience data analysis in the future and may get better results than traditional methods.