The rainfall forecasting models developed in this study are based on the use of ANNs to implement
the pattern recognition methodology. ANNs, which emulate the parallel distributed
processing of the human nervous system, have proven to be very powerful in dealing with complicated
problems, such as pattern recognition and function approximation. It has been shown by
Hornik et al. [l] that an ANN with sufficient complexity is capable of approximating any smooth
function to any desired degree of accuracy. In addition, ANNs are computationally robust, in
the sense that they have the ability to learn and generalise from examples to produce meaningful
solutions to problems even when input data contain errors or are incomplete. A further advantage
of ANNs in relation to short-term rainfall forecasting is that ANNs can be designed to operate
in real-time.