One of the most challenging tasks in atmospheric modeling is accurate spatio-temporal forecasting of extremely severe cyclonic storms (ESCSs). This study critically examined the numerical simulation of ESCS Hudhud, which formed over the North Indian Ocean basin in October 2014 on a medium-range (7-day) scale using six different cloud microphysical schemes of the Weather Research and Forecasting model. NCEP PREBUFR observation was used to improve the initial condition using a three-dimensional variational data assimilation system. A 7-day simulated track, intensity, and rapid intensification of ESCS were validated against the India Meteorological Department best-fit track observations. The least track error was about 42 km, 67 km, 87 km, and 192 km from Day 4 to Day 7 by using the Goddard Microphysical Scheme (MPS). Mean track error for 7 days was about 129 km, followed by the Lin MPS (~ 136 km) and WSM6 MPS (~ 139 km). The least mean absolute error in maximum surface winds is ~ 4 m/s using the Goddard, Lin, and WSM19 MPSs with a correlation coefficient > 0.93. Probability of detection for rapid intensification/dissipation was about 70% using WSM6 and Goddard MPSs, which was relatively better compared to other MPSs. Results clearly demonstrate that latent heating between 500 and 200 hPa was very important and contributed to intensification, and the maximum magnitude of latent heating correlated well with MSW of ESCS. Simulated maximum reflectivity was compared with the Doppler Weather Radar (DWR) observation from Visakhapatnam, and results indicate that reflectivity is well represented. However, the magnitude was slightly under-predicted compared to DWR in the Goddard MPS. The vertical distribution of hydrometeors and microphysical latent heating is found to play an important role in the initiation and development of cyclonic systems. © 2020, Springer Nature Switzerland AG.