This study evaluated the performance of advanced research weather research and forecasting (WRF) modelling system in the forecast of extremely severe cyclonic storm (ESCS) over the Bay of Bengal region that made landfall in Bangladesh on 15 November 2007. Model initial conditions were improved using the three-dimensional variational data assimilation (3D-VAR) system of the WRF model. A set of four numerical experiments were also performed with and without data assimilation techniques at two different grid resolutions (1° × 1° and 0.5° × 0.5°) using NCEP global forecasting system (GFS). Satellite and conventional observations were assimilated using the 3D-VAR modelling system. Results from the assimilation experiment confirms that the analysis using 3D-VAR is better compared to the global datasets in both grid resolutions, attributed due to the availability of quality satellite observations during the analysis period. In addition, the forecasted track, landfall location, maximum wind speed and minimum central pressure of ESCS are better predicted with the data assimilation. The study also presents the forecasted structure in terms of rainfall, wind speed, potential vorticity, and cross-section of horizontal wind speed. Overall, the results clearly revealed that modelling using the 3D-VAR assimilation system is very important in the high-resolution WRF model that enhanced the prediction capability. © 2021, The Author(s), under exclusive licence to Springer Nature Switzerland AG part of Springer Nature.