A Global Seawater Density Distribution Model Using a Convolutional Neural Network.

Area of Science:

• Oceanography
• Geophysics
• Data Science

Background:

• Seawater density is a critical oceanographic parameter influencing gravity field calculations, tidal potentials, and sensor calibration.
• Previous models often use constant density values, leading to significant inaccuracies in oceanographic studies.

Purpose of the Study:

• To develop a comprehensive and accurate global model for seawater density distribution.
• To improve the precision of oceanographic calculations and the design of exploration systems.

Main Methods:

• Utilized a convolutional neural network (CNN) model.
• Trained the model on extensive real-world seawater datasets.
• Incorporated depth, latitude, longitude, and month as input parameters.

Main Results:

• Achieved an absolute mean error and root mean square error below 1 kg/m³ for 99% of test samples.
• The model accurately reflects the global distribution of seawater density.
• Effectively demonstrates the impact of input parameters on density variations.

Conclusions:

• The newly developed global seawater density model offers superior accuracy compared to existing models.
• This model can significantly reduce errors in theoretical ocean models.
• Provides a robust foundation for ocean exploration system design and analysis.