Improvements to the Automated Real-Time Tidal Elevation System

Abstract : Two statistical techniques for making rapid, short-term predictions of the harmonically determined tidal residual are presented. The techniques are applied to an Automated Real-Time Tidal Elevation System (ARTTES) to improve short-term tide prediction. The techniques investigated are conventionally called 'Kriging' and autoregression. The autoregressive method is based on the Yule-Walker equations for stationary time series. Evaluations using both synthetic and prototype data indicate that although Kriging has a slightly larger root mean square error than the autoregressive technique, the autoregressive technique shows bias, while Kriging gives an unbiased estimate.