A genetic neural network ensemble forecast model for local heavy rain

Based on the numerical forecast products of T213 and Japan, a new nonlinear rainstorm prediction model is developed for local heavy rain. The Japanese rainfall forecast products is used to distinguish the likelihood of heavy rain 24 hours later. Then the Chebyshev sliding nested expansion is applied to the forecast field by T213 for forecast factors best correlated with the series of rainfall. And the empirical orthogonal function (EOF) is utilized to select first principal component of different factor groups. Finally, a genetic-neural network forecast model is set up to daily forecasts of the local rainstorms in June-August, 2008. As shown from the model results of the forecast experiment, it is suggested that the model does well in forecasting heavy rain over the Nanning area.