Due to the influence of weak synoptic systems, the relatively weak precursor signals make significant challenges for the forecast of warm-sector rainstorms in southern China. In order to explore the predictability of warm-sector rainstorms in southern China, and further investigate the error growth characteristics of warm-sector rainstorms at different scales, high-resolution ensemble forecast experiments based on the WRF mesoscale numerical prediction model is carried out for a double rainbands precipitation event that occurred in Guangdong province on 30-31 May 2021. The experimental results show that the convergence of strong low-level wind speed is the main convection initiation conditions for this warm-sector rainstorm of southern China, the mesoscale convergence line at the sea-land interface in South China and the strong southwestern boundary layer jet are conducive to promoting convection. The magnitude of forecast errors and their growth rates at different spatial scales in the warm-sector rainstorm are significantly different, and this rainstorm event is less sensitive to small variations in initial perturbation amplitudes. After convection initiation, the error growth exhibits more obvious nonlinear characteristics, the small-scale errors grow rapidly in the form of “upscale error growth” until saturation, and then dominated by mesoscale error growth. The above study shows that the predictability of warm-sector rainstorms in southern China is limited by multiple factors, and the moist convective process can accelerate the growth of mesoscale forecast errors. Moreover, the strong nonlinear characteristics of forecast error growth at different scales in synoptic systems directly limit the predictability.