Predicting Estuarine Salinity from River Inflows

An analysis system is described that enables the second-order and higher nonlinear interactive effects to be estimated in the relationship between the input and output of nonlinear systems. No data are, in principle, required other than an input and output time history over a sufficient period. The method is compared with the functional approach of Weiner and shown to have advantages in practical application. An application to the prediction of salinity in estuaries from river inflows is described. An error analysis for the prediction of salinity at Benicia, Calif., shows that systematic errors in the data and the influence of unaccounted-for variables contribute the bulk of the error in making the predictions and that the error attributable to the analysis leading to the construction of the predictor contributes a smaller error, which is estimated to be on the order of 8% in this application.