Experiments of Bayesian probability flood forecasting based on the AREM model

To improve the accuracy of the flood prediction in forecast period, the Bayesian probabilistic model is introduced to correct the precipitation forecasted by AREM model using the hourly precipitation,flow data of the flood season of Zhanghe Reservoir in Hubei province from 2006 to 2008 and the corresponding 00-60 h hourly AREM forecasting precipitation provided by IHR(Institute of Heavy Rain in Wu-Han).The corrected and non corrected precipitation are analyzed comparatively, and then the both corrected and non corrected forecasting precipitation by AREM model are input into Xin'an-jiang hydrology flood forecast model.The results show that the corrected forecasting precipitation has the higher accuracy than that of non correcte done.The root mean square error was decreased by less than 10%. The average flood deterministic coefficient increases by 10.66%,the average relative error of peak reduces by 3.05% in the verification period.It improved the precision of flood forecast to some extent.