ABSTRACT
This chapter discusses the main elements in the neural network (NN) toolbox; it also addresses the ‘what’ and ‘when’ of NN hydrological modelling. Section 2 contains a brief introduction to the mechanisms and procedures involved-which includes a discussion on architectures and learning; while Section 3 contains a detailed description of the most popular tools that have been used in the field of water related research. Sections 2 and 3 are intended to complement one another and are designed to impart the minimum amount of information that would be required to understand the various operations and processes that are adopted in neurohydrological modelling. There are several respected sources that can be consulted for a more authoritative and comprehensive discussion on generic NN modelling items or issues of interest. Bishop (1995) and Masters (1995) are good academic texts; each book contains a copious amount of in-depth material. Reed and Marks (1999) is oriented towards the developer and practitioner. It describes selected techniques in sufficient detail, such that real-world solutions could be implemented, and technical issues or operational problems could be resolved. Section 4 illustrates the range of different hydrological possibilities and potentials that exist in which to develop and implement a neural solution. Section 5 highlights the numerous opportunities and benefits that are on offer and further strengthens the argument for increased research into the provision of data-driven models. Sections 4 and 5 are thus intended to bolster appeal and to encourage uptake amongst interested parties; the exploration and testing of unorthodox strategies and alternative mindsets can indeed be a rewarding experience that leads to fresh insights and discoveries.
