What type of study is this?

This is a Quantitative Study study.

October 13, 2025Open Access

Comparative Evaluation of AI Models (CNN, GRU, SimpleRNN) for Bathymetric Depth Prediction from Multibeam Echo Sounder Data

Key Points

The GRU model achieved the highest prediction accuracy with RMSE of 0.1961, surpassing CNN and SimpleRNN.
Performance metrics indicate GRU outperformed CNN and SimpleRNN in convergence and capturing depth variations.
This analysis used multibeam echo sounding data extracted in 2023 to assess deep learning model effectiveness.
These findings suggest deep learning techniques can enhance the consistency and efficiency of seabed mapping.

Abstract

This study evaluates the effectiveness of three deep learning models - Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Simple Recurrent Neural Network (SimpleRNN) - in analyzing multibeam echo sounding (MBES) data collected from the southern coastal waters of Vietnam. The objective is to assess the potential of these models in standardizing and improving the accuracy of bathymetric data processing, with a focus on modeling depth variations. Input data were extracted from MBES surveys conducted in 2023, and structured into standardized CSV formats containing longitude, latitude, and depth. Each model was trained and validated using the same dataset, and performance was evaluated using key metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Squared Error (MSE). Results indicate that the GRU model achieved the highest prediction accuracy, with RMSE = 0.1961, MAE = 0.1485, and MSE = 0.0384. In comparison, the CNN model yielded RMSE = 0.2072, MAE = 0.1537, and MSE = 0.0429, while SimpleRNN performed the least effectively. Among the three, the GRU model demonstrated faster convergence and better capability in capturing complex depth variations. These findings suggest that deep learning - particularly GRU - offers a promising alternative to traditional MBES data processing techniques. The approach proposed in this study may help enhance the quality, consistency, and efficiency of seabed mapping, particularly in areas where post-processing resources are limited. Further improvements could be achieved by incorporating environmental variables and uncertainty estimation into the modeling framework. The models were applied to standardize MBES measurements into regular grids, predict seabed depths, and compare the consistency of depth information derived from different sources. Their comparative performance provides practical insights into the suitability of AI models for bathymetric applications.

Read Full Paperexternally

Mark Helpful

Bookmark

Relay

View Full Paper