This study investigates the effectiveness of ensemble singular vectors (EnSVs) for a mesoscale convective system over the East China Sea, a challenge due to its strong nonlinearity. Employing breeding ensembles with varying horizontal resolutions, we compare linear EnSV predictions with nonlinear perturbed forecasts. The results indicate that EnSVs consistently capture sensitivity to synoptic-scale features across all resolutions. However, the growth rate of the EnSVs decreases with increasing resolution. While nonlinear effects become more pronounced at higher resolutions, the nonlinearly developed perturbations from initial EnSVs by full models still demonstrate significant growth in the target region. The nonlinearity is likely associated with mesoscale error growth. These findings suggest that mesoscale EnSVs are capable of identifying key growing modes, but they have limitations in fully capturing the nonlinear amplification of errors inherent in mesoscale phenomena.