Deployment of Lower Cost Water Level Sensors and Machine Learning Models for Local Coastal Inundation Monitoring and Predictions

The frequency of coastal inundation is steadily increasing driven by relative sea level rise. More locations are affected by coastal flooding such as coastal roads, properties along canals, and other infrastructure requiring communities to anticipate and plan for such events. Some locations are nearby tide gauges and benefit from real time and historical measurements. Other locations are further away from the US Water Level measuring backbone and must take into account the local hydrodynamic to accurately predict coastal inundations. The development of lower cost water level sensors is making it possible to measure water levels at many more locations and provide location specific information, although not at the same standard for deployment and maintenance as the stations of the national backbone. Many such sensors have been deployed on the Atlantic coast and installations have started in Texas including a June 2023 installation near Magnolia Beach, Calhoun County, Texas. Differences in the timing and magnitude of storm surges for the data set will be described as compared to the nearby Texas Coastal Ocean Observing Network (TCOON) tide gauges located at Port O’Connor and Port Lavaca.

While measurements are valuable to assess the progression of an event and for long term planning, predictions are even more impactful ahead of potential coastal inundation events. While existing operational hydrodynamic models provide good guidance for regional water level predictions, their resolution may not sufficiently account for the characteristics of some locations such as the local bathymetry that can be intricate and affect significantly the water level dynamic. Machine learning models have proven very effective at predicting water levels for tide stations when sufficient data is available. The talk will present a method to leverage the national backbone of tide stations’ historic data to calibrate a neural network model to predict water levels at Magnolia Beach while accounting for the influence of the strong winds on the local hydrodynamic of the shallow ponds and marshes of this nearshore environment. The machine learning model was first trained to create a synthetic data set of predictions of present and future water levels in the nearby open waters of Matagorda Bay. A secondary ML model was then calibrated to transfer the water level predictions from Matagorda Bay to the new sensor location in a pond near the coastal road most at risk of flooding. The model has been operationalized using the recently deployed Semaphore software (Kastl et al. 2025) and examples during recent events will be shared. Results show the promise of the method to augment flood warning systems available to decision makers with localized information near the infrastructure that they are most concerned during these increasingly frequent coastal flooding events.

“Matthew Aiden Kastl, Texas A&M University - Corpus Christi, Corpus Christi, TX; and B. Estrada Jr., S. L. Stephenson, A. Beasley, F. Tissot, S. Nguyen, S. A. King, and P. E. Tissot Semaphore: A Software Platform for Operational Hosting of Coastal AI Models Presentation at the 15th Conference On Transition Of Research To Operations (15R2O), 2025 AMS Annual Meeting, New Orleans, LS.”