Graduation Semester and Year

Spring 2026

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computer Science

Department

Computer Science and Engineering

First Advisor

Yonghe Liu

Abstract

Maritime shipping plays a important role in supporting the global economy; however, the wakes generated by vessels introduce substantial energy into coastal environments, contributing significantly to shoreline erosion and posing threats to both natural ecosystems and built infrastructure. In shallow coastal zones, these wakes further lead to sediment resuspension and long-term morphological changes. Monitoring such hydrodynamic impacts presents considerable challenges, as conventional techniques such as pressure sensors and acoustic profilers are often expensive, require frequent maintenance, and are prone to data gaps or reduced accuracy under adverse environmental conditions.

This study looks at vessel traffic in Ingleside Bay, Texas, by combining Automatic Identification System (AIS) ship data with camera-based water-level measurements. It measures the depression wakes left by ships and examines how they influence sediment movement through a wave–sediment interaction framework. By pinpointing when the forces from passing vessels exceed the natural strength of local shorelines, the study highlights “dangerous” speed–size combinations that increase the risk of erosion. The results will help define speed–size thresholds for safer traffic management, validate them with field data, and compare their performance against existing empirical models. With the addition of deep learning techniques, this research moves toward building accurate, scalable tools for predicting vessel wakes and supporting evidence-based shoreline management in coastal waterways. This dissertation proposes a novel three-phase system grounded in artificial intelligence. In the first phase, a detection method based on the YOLO framework is developed to accurately measure water level changes from staff gauge imagery under diverse environmental conditions. The second phase introduces a predictive framework based on the Interpretable Multi-Variable Long Short-Term Memory (IMV-LSTM) model to predict water level dynamics during documented ship events. This approach not only forecasts water level fluctuations but also quantifies the influence of key variables such as ship size, ship speed, and timing through variable-wise attention mechanisms. The third phase extends the model to address complex real-world conditions involving the simultaneous passage of multiple vessels with varying sizes and speeds. Beyond forecasting, this phase defines safe transit conditions across different regions of the navigational map, identifying areas where vessels of specific size and speed classes can operate without inducing hazardous drawdown or shoreline erosion. In doing so, the framework fills a critical research gap by enabling robust, real-time decision support in scenarios where traditional empirical approaches are inadequate.

Disciplines

Digital Communications and Networking

Download

Available for download on Monday, May 10, 2027

Share

COinS