Sea Water Level Estimation Using Six Different Artificial Neural Networks Training Algorithm

Abstract

Water level estimation is important at various time intervals using the records of past time series in water resources engineering and management. For instance, sea level affects groundwater tables in low-lying coastal areas, Also hydrological regimes of some coastal rivers. Therefore, a reliable forecast of sea-level variations is required in coastal engineering and hydrologic studies. In this study, it has been tried to predict the changes in sea level by six different artificial neural networks (ANN’s) training algorithms (Quasi-Newton, Conjugate Gradient, Levenberg-Marquardt, One Step Secant, Resilient back propagation and scaled conjugate gradient algorithms) and multiple linear regression (MLR) methods, three time steps, for a set of time intervals comprising 6 hour, 12 hour, 18 hour, 24 hour, 2 day time intervals using observed sea levels. The measurements from a single tide gauge at Hillarys Boat Harbor Western Australia. The results of the ANN’s algorithms are compared models with respect to root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R2). The comparison results indicate that the Levenberg-Marquardt is faster and has a better accuracy than the other training algorithms in modelling sea level. The Levenberg-Marquardt with RMSE = 0.004 m, MAE = 0.002 m and R2 = 0.999 in test period was found to be superior in modelling sea level than the other algorithms, respectively.