ABSTRACT

NN rely on the provision of adequate data sets, and where these are available, NN may be programmed to search for patterns within the data. On the basis of this patternmatching, forecasts are made on independent data sets first for model validation and then for operational purposes. NN are one approach within the broader hydroinformatics framework which emerged in the 1990s as a route to managing information overload in an effective way (Govindaraju & Rao, 2000). Price (2000) recognises four strands to hydroinformatics, the mathematical and physical sciences understanding, the handling of data and the human cultural element. One of the significant strengths of the NN approach is that it can handle all data types.