Land–atmosphere energy and moisture exchange can strongly influence local and regional climate. However, high uncertainty exits in the representation of land–atmosphere interactions in numerical models. The parameterization of surface exchange process is greatly affected by varying the parameter Czil which, however, is typically set to a domain-wide constant value. In this study, we examine the sensitivity of regional climate simulations over China to different surface exchange strengths using three Czil schemes (default without Czil , constant Czil = 0.1, and dynamic canopy-height-dependent Czil -h schemes) in the 13-km-resolution Weather Research and Forecasting model coupled with a Noah land surface model with multi-parameterization options (WRF/Noah-MP). Our results demonstrate that the Czil -h scheme substantially reduces the overestimations of land–atmosphere coupling strength in the other two schemes, and comparisons with the ChinaFLUX observations indicate the capability of the Czil -h scheme to better match the observed surface energy and water variations. The results of the Czil schemes applying to four typical climate zones of China present that the Czil -h simulations are in the closest agreements with the field observations. The Czil -h scheme can narrow the positive discrepancies of simulated precipitation and surface fluxes as well as the negative biases of Ts in areas of Northeast, North China, Eastern Northwest, and Southwest. Especially, the above remarkable improvements produced by the Czil -h scheme are primarily over areas covering short vegetation. Also noted that the precipitation simulated by the Czil -h scheme exhibits more intricate and unclear changes compared with surface fluxes simulations due to the non-local impacts of surface exchange strength resulted from the fluidity of the atmosphere. Overall, our findings highlight the applicability of the dynamical Czil as a better physical alternative to treat the surface exchange process in atmosphere coupling models.