Based on the Bayesian model for decision supporting, the decision making procedures in the meteorological service for the Beijing 21 July 2012 extra torrential rain are simulated and analyzed in this paper. The diagnosis in decision makings on the typical torrentail rain event try to prove its possibility of application in supporting decision making model (SDMM) to attain the optimal decision in solving the uncertainties problems of decision making, under the current condition of data resources (e.g., the mesoscale NWP system and observation data)， weather forecasting and meteorological observing technological levels. With the NWP products on that day, the floods and debris flow risks both have been assessed to be high. Combine with the data of floods threshold and rainfalls recorded in the historical flood events and the predicted rainfall magnitude distribution on that day, a high torrential rain risk can also be recognized consequently. The posteriori probability deduced using the Bayesian model is only 23.1%. However, considering the expected losses (EL) gap in predicting severe weather and non-severe weather, the non-severe weather prediction EL can be obviously greater than the severe weather prediction EL. Therefore, the optimal decision making in that situation would have been to publish severe weather warning and pick the pessimistic scheme in the meteorological service advisably. The simulation of the meteorological service on the extra torrential rain day reveals that the tolerability to the severe weather forecasting and warning uncertainty can relieve the pressure on forecasters who are often afraid of giving false forecasting and warning. In additional to perfect emergency preparations, the uncertainty information could be published, and properly be delivered to the actual meteorological information users in weather forecasts, which ultimately helps to converte the deterministic weather forecasting mode into that of non-determinacy.