Abstract: The weak precursor signals resulting from the influence of weak synoptic systems pose significant challenges to the forecasting of warm-sector heavy rain in South China. To explore the predictability of warm-sector heavy rain in South China and further investigate the error growth characteristics of warm-sector heavy rain at different scales, high-resolution ensemble forecast experiments based on the WRF (Weather Research and Forecasting) mesoscale numerical prediction model are performed for a double rainband 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 primary condition for convection initiation in this heavy rain. Additionally, the mesoscale convergence line at the sea–land interface in South China and the strong southwesterly boundary layer jet are conducive to the development and enhancement of convection. The magnitudes of the forecast errors and their growth rates at different spatial scales in the warm-sector heavy rain are significantly different, and this heavy rain event is less sensitive to small variations in initial perturbation amplitudes. After convection initiation, error growth exhibits pronounced nonlinear characteristics, with small-scale errors rapidly amplifying in the form of “upscale error growth” until reaching saturation, beyond which mesoscale errors become dominant. The results of this study indicate that multiple factors limit the predictability of warm-sector heavy rain in South China and that a moist convective process can accelerate the growth of mesoscale forecast errors. Moreover, the strong nonlinear characteristics of the forecast error growth at different scales in synoptic systems directly limit predictability.