River Stage Forecasting Using Artificial Neural Networks

Abstract

Methods to continuously forecast water levels at a site along a river are generally model based. Physical processes influencing occurrence of a river stage are, however, highly complex and uncertain, which makes it difficult to capture them in some form of deterministic or statistical model. Neural networks provide model-free solutions and hence can be expected to be appropriate in these conditions. Built-in dynamism in forecasting, data-error tolerance, and lack of requirements of any exogenous input are additional attractive features of neural networks. This paper highlights their use in real-time forecasting of water levels at a given site continuously throughout the year based on the same levels at some upstream gauging station and/or using the stage time history recorded at the same site. The network is trained by using three algorithms, namely, error back propagation, cascade correlation, and conjugate gradient. The training results are compared with each other. The network is verified with untrained data.