ABSTRACT
Following a successful pilot project to demonstrate the application of artificial neural network (NN) methods to flood forecasting (Openshaw et al., 1998), the main barrier to the practical implementation and use of NN based flood forecasting systems was considered to be one of communication. In order for the river engineers and hydrologists who are responsible for flood warning and control to place their trust in such tools, the world of NN must first be demystified. To help address this issue, an integrated NN based flood forecasting system, called RLF/1, was developed using web-based technologies. RLF/1 comprises a suite of software tools running on a SUN-Solaris workstation. The tools have been written in C++ and are integrated via HTML and UNIX shell scripts. The system can be accessed and run on-line from the following address: https://www.ccg.leeds.ac.uk/simon/intro.html
The aim of RLF/1 is to demonstrate that NN solutions are a practical tool for realworld flood forecasting operations. The Internet offers a system that can be openly accessed at minimal cost by engineers and scientists from around the world who can experiment with the NN models at their leisure. Time series data can be uploaded and a real-time forecast simulated using stored NN models. New models can also be constructed, trained on-line and then used to provide simulated forecasts. The use of web-
based technologies means that background material, on-line help, instructions and links to relevant information are fully integrated within the software package.
