Flood Early Warning System Develop at Garang River Semarang using Information Technology base on SMS and Web

One of large rivers flowing in Semarang City is Garang River with watershed area about 203 km ²and having characteristics: high flood discharge and flash flood. The flash flood on January 25^th1990 caused more than 45 people died and lost 8.5 billion rupiahs. In this research, a flood early warning system made, anautomatic rain fall recorder (ARR) was placed in upstream of Garang River and an automatic water level recorder (AWLR) placed in Simongan weir. To receive rainfall and water level data, a computer as server placed in Semarang government office. This functions for sending flood status (prepared/caution/careful) to flood attendants, stake holders. Artificial neural network (ANN) used to predict water level where the input are rainfall in upstreamriver for 1, 2, 3, 4 or 5 days before and the water level in the down stream for 3 hours before the occasion. The output water level in down stream for next 2 hours. This system integrated with information technology that is SMS (Short Message Services) and Web that the flood early warning can be accessed anywhere as long as communication network is available. The result shows time for sending data of rainfall and water level telemetry system less than 10 minutes, while information system built to give flood early warning information less than 10 minutes. Consequently it fulfill in which the time to receive the information less than time when flood happened in 2 hours. The result of optimum predicting during the ANN training is model 4, 20 neurons speed training 0.9, momentum 0.3 which input rainfall are four days before the occasion and the water level are 3 hours before having MSE 0.0046. Finally, the results of survey to flood attendants and staffs from government show that 86% of the respondent absolutely need the flood early warning system.