Neural Network Surge Predictions: Design, Implementation, History, and Performance Comparison

Accurate water level forecasts are important for navigation, safety ahead and during coastal flooding and coastal works in general. In particular, the need for accurate coastal flood watches and warnings is growing due to increased nuisance flooding frequencies associated with relative sea level rise. While astronomical forcing (tides) is well tabulated, water level changes can be dominated by meteorological factors and other oceanographic forcings such as strong winds and changes in longshore currents, which are unaccounted for in tidal predictions. In coastal areas such as the Northwest Gulf of Mexico which is microtidal and exposed to high winds, the impact of these meteorological factors is often greater than the tidal range. Artificial Neural Network (ANN) models were developed and implemented in the early 2000’s, (Cox et al. 2002, Tissot et al., 2001, 2003, 2004) to predict short-term water levels, up to 48 hours, based on previously observed water levels, wind speeds, wind directions, and forecasted tidal and meteorological data. Forecasted meteorological data were extracted from successive versions of the North American Model (NAM) through a collaboration with the local office of the National Weather Service. The ANN models were implemented for several locations along the Texas coast with observational data collected as part of the Texas Coastal Ocean Observation Network (TCOON) and the National Water Level Observation Network (NWLON). The design, implementation and performance of the models will be described including examples of the use of the predictions. The method was also tested for other locations in the Gulf of Mexico and along the Atlantic coast with different results. Finally the performance of the method will be compared with other modeling techniques for the NWLON station of Corpus Christi, Texas, for a set of recent high water level events that took place starting in the summer of 2015 through the summer of 2017.

Tissot, Philippe E., Daniel T. Cox, and Patrick Michaud. "Neural network forecasting of storm surges along the Gulf of Mexico." In Ocean Wave Measurement and Analysis (2001), pp. 1535-1544. 2002.

Cox, D.T., Tissot, P. and Michaud, P., 2002. Water level observations and short-term predictions including meteorological events for entrance of Galveston Bay, Texas. Journal of waterway, port, coastal, and ocean engineering, 128(1), pp.21-29.

Tissot, P.E., Cox, D.T. and Michaud, P.R., 2003, February. Optimization and performance of a neural network model forecasting water levels for the Corpus Christi, Texas, Estuary. In 3rd Conference on the Applications of Artificial Intelligence to Environmental Science, Long Beach, California.

Tissot, P., Cox, D., Sadovski, A., Michaud, P. and Duff, S., 2004. Performance and comparison of water level forecasting models for the Texas ports and waterways. In Ports 2004: Port Development in the Changing World (pp. 1-10).