Application of nonlinear system identification to the Lower Mekong River, Southeast Asia

Abstract

The generalized functional series representation of systems was solved numerically by using a transformation of input that considerably reduces the computational difficulties and a multiple‐regression analysis that establishes an optimum model. Computer programs for constructing such models were applied in the construction of two models for the lower Mekong River. Both models have single outputs, the daily gage height of the river at Chaudoc, South Vietnam. One model is of the single‐input type, in which the input is the net average daily discharge below Phnom Penh, Cambodia. The second model is of the double‐input type; it has as a second input the daily rainfall at Takeo, Cambodia. A total of 1497 daily measurements from January 1, 1964, to February 5, 1968, were used in the analysis. The predicting capability of the technique was tested by using the model constructed for 1104 days as a predictor for the remaining 393 days.