| Abstract | The flood situation in Thailand is one of issues that affect the quality of life for Thai people. Chao Phraya River Basin has 8 tributaries; Ping, Wang, Yom, Nan, Sakaekrung, Ta Chin, Main Chao Phraya and Pasak River Basin. These tributaries frequently cause damages to public and private properties in the flood plain areas especially during flood seasons. The damages downstream of the Pasak River are due to the back water effect from Chao Phraya River in flood seasons. The Royal Irrigation Department (RID) started to conduct a feasibility study on the Pasak Storage Dam Project in 1965. On 19 February 1989, His Majesty the King Bhumibol graciously reviewed the project and suggested that Dam should be constructed to help ease water shortages problem in farm areas in the Pasak River Basin and control floods in that basin, Bangkok and its vicinity. As suggested, the Pasak Storage Dam and other related works were constructed from 2 December 1994 and were completed in 30 December 1999. Telemetering system is efficiently one of the sustainable tools of operation and management in the Pasak Jolasid Dam (PJD). At present, this system is very useful for measuring, transferring, analyzing and monitoring flood situation information by forecasting flood with some mathematical models. To increase operation and management efficiency, the PJD had installed telemetering system which composed of a master station at the PJD, sub-master station at the regional irrigation office 10, Lopburi province, and 12 remote stations. All stations collect rainfall and streamflow data with real time data collection. In actual practice, the operator uses the average historical rainfall to predict upcoming rainfall daily based on his intuitive experience and without using any model. But such forecasts often do not work out well in many cases. For successful operation management and sustainability of the PJD, a real time daily forecasting is a tool that can make the decision easier for the operator of the dam. In this study, the Neural Networks (NNs) is chosen to forecast daily rainfall at the representative rain station in the upstream of the PJD. The NNs architecture giving the best performance is chosen based on the period of training and testing phases. The meteorological variables used as input in the model are total daily rainfall, relative humidity, temperature, cloudiness, pressure, and weather forecast code. The sensitivity of inputs influencing the performance of NNs is analyzed. The rainfall output from NNs is used as an input for MIKE 11 with NAM and 1-ID module for forecasting streamflow. The forecasted streamflow is used to find out the occurrences of flood in the flood plain area. The forecasted daily rainfalls are compared with the observed data by using the historical and real time data of rainfall. It has been found that the efficiency index is satisfactory and then the model can perform forecasting flood very well. For improving the accuracy of flood forecasting, radar images or satellite images data should be used in future |