Abstract: Warm-sector rainstorm events over MLRYZ (middle-lower reaches of the Yangtze River) usually exhibit large prediction uncertainties because they are easily affected by the complex underlying surface, particularly in the convection initiation stage. In this paper, high-resolution numerical and convection-permitting ensemble simulations are performed on a warm-sector rainstorm that occurred over MLRYZ on 23 June 2020, that was affected by complex terrain to investigate the trigger mechanism and reveal the limited predictability of this event. The convection initiation stage is analyzed using air parcel Lagrangian backward trajectory analysis, sensitivity experiments for removing terrain and closure thermal effects, and ensemble sensitivity analysis. Results show that the lifted air parcels mainly come from the planetary boundary layer below 1.5 km. Valley wind driven by the thermal effect of the Xianxia and Shan Mountains is a dominant dynamic source that triggers local convergence and lifting. The divergence of high and low levels, the vertical configuration of moist potential vorticity, and dipole potential vorticity anomaly are good indicators of convection initiation. In addition, this event is highly sensitive to the 2-m temperature and apparent heat source at a low level, indicating the importance of the accurate underlying forcing for the warm-sector rainstorm prediction. The initial condition sensitivity experiments with gradually decreasing initial errors suggest that the predictability of warm-sector convective events is considerably lower than that of northern frontal events. The RMDTE (root mean difference total energy) of frontal convections can decrease continuously with a reduction in initial errors, while the RMDTE curves of warm-sector convection still reach the original level, showing nonlinear convergence characteristics. Therefore, data assimilation techniques may be prioritized for frontal convection with strong synoptic forcing to reduce initial errors and further decrease forecast errors. However, more attention should be paid to ensemble forecasts to represent uncertainties in the convection initiation process of warm-sector rainstorms under complex terrain.