This study focuses on the sensitivity of tropical cyclones (TCs) simulations to physics parametrization scheme for TCs in the Bay of Bengal (BOB). The goal of this study was to arrive at the optimum set of schemes for the BOB region to increase forecast skill. Four TCs, namely Khaimuk, Laila, Jal and Thane have been simulated through the weather research and forecasting (WRF) model with all the physics parametrization schemes available in WRF, and the optimum set of schemes is arrived at. The analysis shows the cumulus, microphysics and planetary boundary layer parameterizations exert a very significant influence on the TC simulations than land surface, short-wave radiation and long-wave radiation parameterizations. With this optimum set of physics schemes, the impact of assimilation of National Centers for Environmental Prediction Automatic Data Processing upper air observations data in the TC simulations has been studied by using three-dimensional variational (3DVAR) data assimilation technique. The control run (without assimilation) and the 3DVAR-simulated tracks and maximum sustained wind speed have been compared with the Joint Typhoon Warning Center observed tracks and wind data. The model-simulated precipitation is validated with Tropical Rainfall Measuring Mission 2A12 surface rain rate and 3B42 daily accumulated rain data. Bias score and equitable threat score have been evaluated for both instantaneous rain rate and 72-h accumulated rain. © 2015, Springer Science+Business Media Dordrecht.