In order to study sensitivity of physical parameterization schemes (PPSs) in the WRF model on the simulation of landfall tropical cyclones (TCs), we design 43 sensitivity experiments based on three types of PPSs, and select three disastrous TCs that landed on the east coast of China, namely Lekima (2019), Hagupit (2020) and In-Fa (2021). Qualitative analysis and quantitative evaluation are carried out from the aspects of model simulated TC track and intensity, as well as the landfall position, time and intensity. The 72h-average track errors of all sensitivity experiments are 105 km, 85 km and 38 km, by three TCs respectively. The temporal variations of simulated TC intensity in all sensitivity experiments are highly consistent with the observations, but the simulated TC intensities are inevitably weaker than the observations. The results of three groups of experiments show that the track and intensity of landfall TCs are most sensitive to cloud microphysical parameterization scheme (CMPS), but insensitive to planetary boundary layer parameterization scheme (PBLPS). All the simulated landfall positions are more southward than the observations, and the intensities at landfall moment are underestimated compared with the observations. The TC landfall position is the most sensitive to cumulus convection parameterization scheme (CCPS), followed by PBLPS and CMPS. However, the situation for the TC landfall intensity is the opposite. Finally, some optimal PPSs for simulating TC track, intensity and landfall position are given, so as to provide some references for the simulation of landfall TCs in regional prediction systems.