Abstract: The real time forecasting experiments during the main flood season in 2003 are performed two times a day by two model systems,with one the mesoscale heavy rain forecasting model system named AREMS(Advanced Regional Eta-coordinate Model System) including three-dimensional variational assimilation and the other the AREM model version 2.1.The 24hr accumulated precipitation forecasts are verified in different parts of China.The results show as follows:(1) AREMS has a better forecasting ability for summer precipitation in China,with the highest Ts score in the middle-lower valleys of Yangtze River.Fault-hitting rates are bigger than no-hitting rates in the middle-lower reaches of Yangtze River,North China and South China and it is just reversal in Northeast China and the east part of Southwest China.(2) The changes of model conditions(including model level top,lateral boundary condition and initial condition) have different impacts on the 24hr forecasted precipitation in different areas,with Northeast China and North China influenced most evidently from June to August,South China and the east part of South-west China coming to the next and the middle-lower reaches of Yangtze river least impacted.(3) The changes of model conditions have different influences on Ts scores in different areas.In general,AREMS has most obviously improved precipitation forecast in Northeast China,especially for the period from 12 to 36 forecasting hours.The improvement in North China is also obvious.However,the precipitation prediction ability in South China is generally decreasing,with little improvement in the middle-lower valleys of Yangtze River and the east part of Southwest China.(4) The changes of model conditions have different effects on false-hitting rates and no-hitting rates in different areas.The latter ratios are obviously reduced in North China as opposed to false-hitting rates.Most no-hitting and false-hitting rates in Northeast China are less than those of AREM2.1.In the east part of South-west China,no-hitting rates are slightly reduced at 08h as opposed to false-hitting rates,but it is reversed at 20h.(5) There exist obvious differences of Ts scores in the middle-lower valleys of Yangtze River,South China and the east part of South-west China in June,July and August.