NEURAL NETWORKS IN RECONSTRUCTING MISSING WAVE DATA IN SEDIMENTATION MODELLING

Decision support systems for efficient water management and control require the measured time series data such as rainfall, runoff, water level, waves, wind etc. In practical situations these time series often contain gaps that might originate from failures of apparatus or some other reasons. For example the wave data measured at the wave measuring station Europlatform in the North Sea consists of gaps in the wave data time series. This data is of paramount importance for understanding the sediment dynamics in the North Sea along the Dutch coast in general and for assessing the sedimentation in the nearby approach channel of the Port of Rotterdam in particular. An artificial neural network (ANN) has been used for filling in the gaps of the wave data time series from the wind data time series. Average mutual information has been used for determining the lag effect of wind. The direction of wind data has been encoded with a circular coding algorithm. The accuracy of the ANN model in estimating wave heights was found to be reasonable and the model was used to fill in the gaps of the wave data time series for the purpose of subsequent sedimentation modelling, also based on an ANN.