OPTIMISATION OF NEURAL NETWORKS FOR RAINFALL-RUNOFF MODELING

Abstract

The relationship between rainfall and runoff is one of the most complex hydrologic phenomena to comprehend due
to the tremendous spatial and temporal variability of watershed characteristics and precipitation patterns, and the number of
variables involved in the modeling of the physical processes. As a result of these difficulties, and of a poor understanding of
the real-world processes, empiricism can play an important role in modeling of R-R relationships. Artificial Neural
Networks (ANNs) are typical examples of empirical models. Their ability to extract relations between inputs and outputs of a
process, without the physics being explicitly provided to them, theoretically suits the problem of relating rainfall to runoff
well, since it is a highly nonlinear and complex problem. The goal of this investigation was to develop rainfall-runoff models
for the river Jhelum catchment that are capable of accurately modelling the relationships between rainfall and runoff in a
catchment. Two types of ANN models viz. Back Propagation networks (BPN) and Radial Basis function (RBF) were
developed. The network architecture in the back propagation network was changed by changing the number of neurons in the
hidden layer. The analysis of performance of the various models was carried out by statistical analysis technique .The
comparison was based on various statistical parameters like root mean square error (RMSE) and R2
.