Artificial Neural Network Model Application on Long Term Water Level Predictions of South Florida Coastal Waters

Abstract

Increasing the number and quality of water level data is very important to the hurricane and surge research. Long-term water level data of local stations in estuaries and inland waterway can be used to validate the performance of traditional storm surge models in complex coastal environments. However, field data collections are expensive and often limited by available research budget. Only a few water level stations operated by NOAA provide long-term observations close to 100 years. In this study, a Feed-forward backpropagation ANN Model has been applied to establish quantitative relationship between short-term water level measurements at Naples station and long-term water measurements at Cedar Key station. Using water levels at NOAA stations, the neural network model can be used to derive reliable long-term historical water level data at other stations along south Florida coast by model training and verification. Long-term water level data derived from the ANN model can be used analyze historic hurricane surge hydrograph in Florida coast after removing tidal signals. The data can also be used as boundaries for modeling hurricane storm surges in bays, estuaries, and coastal waterways.