抄録
The severe rainfall prediction method using artificial intelligence which simulates the act of weather forecast experts who are well versed in the local weather is proposed in order to support the flood control.
Under the condition that the system does not take the place of the numerical weather forecast, the objectives of development of the system are as follows; i) to forecast severe rainfall in fine grid scale by consideration of subgrid phenomena which are difficult to be expressed in numerical models, and ii) to make a real-time explanation of the important causes of severe rainfall to river managers.
Then, the subsystems of the SRAI are developed and discussed in terms of their performance. The accuracy of the system was 80% for forecast of two hours ahead rainfall over 10mm/hr. The system shows the important initial conditions for severe rainfall effectively using backward reasoning.
