Abstract: Using ground based precipitation data, satellite cloud images, radar echoes and NCEP reanalysis data, this study analyzes characteristics of MCSs(Mesoscale Convection Systems)associated with a continuous rainstorm that occurred over the coastal areas of South China within a monsoon trough, and investigates the possibility of numerical prediction of the MCSs precipitation. The analysis shows that the rainstorm was caused by a number of successive MCSs. In similar environments, these MCSs developed into different modes with different horizontal scales. The maximum could be organized into a meso-α scale MCC, while meso-β scale linear or banded convective systems were more common. Comparative analysis of two MCSs that respectively presented as quasi-circular MCC (MCS-2) and banded MCS (MCS-4) reveals that convection in both of the two MCSs began to develop at the nighttime, showing a good correlation with fluctuations in the southerly low-level jet in the monsoon trough. Examination of the observed proximity soundings shows that it is difficult to distinguish the modes and the evolutionary tendencies of different MCSs using diagnostic variables such as convective available potential energy (CAPE) and vertical wind shear, which are considered to be good precursors for MCSs precipitation prediction. The representativeness of soundings used to derive these quantities will affect the application of various forecasting tools such as that based on the "ingredients-based method". Prediction simulation with the GRAPES_GZ model shows that it is possible to explicitly predict MCSs precipitation by the numerical model. In comparison, the 3 km resolution run can better predict the occurrence of MCSs-induced rainstorm, but are sensitive to the option with or without CP (convective parameterization) scheme. Though running without the CP scheme could better predict earlier stage rainfall in both MCSs, it overpredicted precipitation as the MCSs evolved into their mature stages. How to describe mechanisms for organized convection in MCSs and how to deal with problems of CP schemes at "grey-zone resolution" are two issues that need to be considered carefully.