The importance of hydrological variables and the complexity of hydrological phenomenon lead to the necessity of hydrological modelling in a water resources management system. In the recent years, data-driven modelling approach is emerging as a new methodology. This approach is based on identification of the relationships between input and output variables of a hydrological system based on the historical data. In this context, the purpose of this study was to investigate the applicability of different data-driven modelling techniques in modelling hydrological variables, namely: artificial neural networks (ANN) and model trees (MT) for runoff modelling nearest neighbour approach and fuzzy logic approach for error assessment classification techniques for classifying runoff chaos theory for rainfall prediction. The applicability of these techniques are studied by doing a case study for a catchment named Sieve catchment located in Italy, which has a catchment area of 822 square kilometres and response time of about six hours. The result shows that both ANN and MT produce excellent results for one hour ahead prediction of runoff, acceptable results for three hours ahead prediction of runoff and unsatisfactory result for six hours ahead prediction of runoff. The comparative performance of ANN and MT shows that both techniques has almost similar performance for one hour ahead prediction of runoff, but the result of ANN is a bit better for higher lead times. Further, the uncertainty of prediction was analysed. In the next phase, the applicability of classification methods for runoff prediction was investigated (with the runoff events assigned as low, medium and high); these techniques appeared to be robust and accurate, and the result can be useful for management purposes. Finally, the analysis of non-linear dynamics of rainfall shows that the rainfall is a chaotic time series, which can be predicted to a certain degree of accuracy for short lead times. It can be concluded that simpler and easier modelling approach like data-driven approach can provide quick and cost effective solution for hydrological problems in such cases where the main concern is the prediction of the magnitude of variable rather than explicitly dealing with the complex system.
Back to the list of MSc abstracts