For the heavy rainfall event in South China in 19-20 May 2015, ECMWF-IFS model (EC mo-del) overestimated its rainfall intensity near the large-scale shear line, but underestimated the heavy rainfall induced by the mesoscale convective systems (MCSs) in warm sector, resulting in a northward displacement bias of the forecasted rainfall compared with the observation in Guangdong Province on 20 May 2015. In this paper, a high resolution numerical simulation of WRF model is performed to explore the source of forecasting error of EC model. The results indicate that the well-organized MCSs in the warm sector induced a significant cold pool outflow, which converged with the strong warm and moist southwest flow. New MCSs were triggered along the convergence line continually and produced heavy rainfall in the warm sector. The WRF model successfully depicts the whole process, while EC model failed to present the above mechanism and caused the underestimated rainfall in the warm sector. The feedback of convective rainfall to synoptic scale flow in South China can be described by WRF model, which can simulate the well-orga-nized MCSs. Most of the rainfall in EC model is the stratiform rainfall caused by the shear line, which may further strengthen the circulation in the middle and low level and increase the precipitation along the shear line in turn. The underestimate of convective rainfall in warm sector and a strong stratiform rainfall feedback work together to cause a northward displacement bias of the forecasted rainfall in EC model.