Search

Water current modelling and prediction techniques along coastal inlets have attracted growing concern in recent years. This is largely so because water current component continues to be a major contributor to movement of sediments, tracers and pollutants, and to a whole range of offshore applications in engineering, environmental observations, exploration and oceanography. However, most research works are lacking adequate methods for developing precise prediction models along the commodore channel in Lagos State. This research work presents water current prediction using Artificial Neural Networks (ANNs). The Back Propagation (BP) technique with feed forward architecture and optimized training algorithm known as Levenbergq-Marquardt was used to develop a Neural Network Water Current Prediction model-(NNWLM) in a MATLAB programming environment. It was passed through model sensitivity analysis and afterwards tested with data from the Commodore channel (Lagos Lagoon). The result revealed prediction accuracy ranging from 0.012 to 0.045 in terms of Mean Square Error (MSE) and 0.80 to 0.83 in terms of correlation coefficient (R-value). With this high performance, the Neural network developed in this work can be used as a veritable tool for water current prediction along the Commodore channel and in extension a wide variety of coastal engineering and development, covering sediment management program: dredging, sand bypassing, beach-contingency plans, and protection of beaches vulnerable to storm erosion and monitoring and prediction of long-term water current variations in coastal inlets. [ABSTRACT FROM AUTHOR]

In this work, 500 hourly water level tidal data were used to perform least squares tidal harmonic analysis. Eleven tidal constituents were used for the harmonic analysis. Astronomical arguments (v + u) and the nodal factor (f) were computed for each tidal constituent and at each observational period with a programme written in Matlab environment. The harmonic constants determined from the least squares tidal harmonic analysis were substituted into a tidal prediction model to predict hourly tidal data and tidal predictions at 5 minutes' intervals. Series of high and low water heights from the tidal predictions made at 5 minutes' intervals were determined and matched with their corresponding times. Autocorrelation at lags 1 to 30 for the residuals of the observed and predicted tidal data shows that there is no significant correlation in the range of the 30 lags. The series of residuals of the observed and predicted tidal data is therefore white noise. [ABSTRACT FROM AUTHOR]