Enhancing coastal resilience: AI-driven seasonal to multi-year water level predictions for the Texas Gulf Coast

The ongoing rise in sea level over the last decade has substantially increased the frequency of coastal inundation episodes across regions such as the Texas Gulf coast. Nevertheless, presently, stakeholders lack the essential tools for adequate preparation in response to these events. To address this gap, our research introduces a multilayer perceptron model specifically designed to forecast water levels across various time scales, ranging from months to years. This model uses a unified signal derived from a refined analysis of historical tide gauge time series along the Texas coast. Our model goes beyond existing short-term and location-specific predictions by offering forecasts that span from seasonal to multi-year timescales, and cover the entire coastline. This extended outlook, provides coastal stakeholders and beach managers with the lead time needed to better plan and implement effective mitigation measures. This cutting-edge model, which could be adapted for other coastal regions, achieves a mean absolute error of less than 8 cm – a substantial improvement over the existing regional prediction tools, which have a variability of approximately ±15 cm. • ML model predicts seasonal to multi-year water levels on the Texas Gulf Coast. • MAE < 8 cm, improving over ±15 cm in current stakeholder forecasting tools. • Seasonal cycle analysis reveals drivers of water level variability. • Framework is adaptable to other vulnerable coastal regions worldwide. • Model aids beach managers with real-time early warnings for inundation response.