Impact of Scale-Dependent Optimal Growth Initial Errors on Mesoscale Predictability of the “21.7” Zhengzhou Extreme Rainfall Event

Abstract This study applies the conditional nonlinear optimal perturbation (CNOP) method to investigate the impact of optimally growing initial errors at different scales on the mesoscale predictability of the Zhengzhou extreme rainfall on 20 July 2021 (0000–2400 UTC) and explore the intrinsic predictability by varying the initial uncertainty constraints. Using a convection-permitting model with 3-km resolution developed by the China Meteorological Administration, particle swarm optimization was employed to identify CNOPs under different constraints. These CNOPs, representing the largest-growing initial errors under specific constraints, were further categorized into all-scale CNOP (A-CNOP), larger-scale CNOP (L-CNOP, ≥200 km), smaller-scale CNOP (S-CNOP,