Application of artificial neural networks and M5 model trees to modeling stage-discharge relationship

摘要

A reliable estimation of discharge in a river is essential to establish an efficient surface water planning. Hydrologists use historic data to establish a relationship between the stage and discharge, which is known as a rating curve, specific to the location of measurements. Once a relationship is established it can be used for predicting discharge from future measurements of stage only. Unfortunately, the relationship between stage and discharge is not always unique and often exhibit random fluctuations. The recent advances in the data-driven modelling techniques suggest that these techniques may be utilised in modelling the complex relationship between stage and discharge. In the present research data-driven model of the stage-discharge relationship is built with an M5 model tree using data of a stage-discharge measuring station. The predictive accuracy of this model is compared with an ANN model and a conventional rating curve built with the same data. It is concluded that the models built with the data-driven modelling techniques show superiority in predicting the discharge over the conventional model.