A detailed assessment was carried out for extremely severe cyclonic storms (ESCSs) that developed over the Bay of Bengal during 1990–2020 using the India Meteorological Department (IMD) best-fit track data. During the past 30 years, the maximum intensity of land-falling ESCS has increased 26%, which is an average of about 8% increase per decade. Analysis signifies an increasing trend in the life cycle, duration of ESCS stage, and maximum wind speed of land-falling ESCSs. It is observed that the stage of land-falling ESCSs was very severe, and therefore, reliable forecast of land-falling ESCSs is very important having wide socioeconomic implications. Furthermore, a case study investigated the performance of Advanced Research of Weather Research and Forecasting (ARW) model for a 6-day forecast of ESCS Hudhud that developed over the Bay of Bengal during 2014. Performance evaluation was based on the impact of model domain size, vertical resolution, data assimilation, sea surface temperature (SST), gravity wave option (GWO), and air-sea flux parameterization schemes. Initial condition was improved through WRF 3D variational data assimilation (WRF-3DVAR) system. Thereafter, the simulated track, intensity, and intensification of the storm were validated against available IMD best-fit track datasets. Results from the ARW model indicate that the track and intensity of ESCS Hudhud were influenced by domain size, vertical resolution, data assimilation, SST, GWO, and air-sea flux schemes. Study deciphers that bigger domain, higher vertical resolution, data assimilation, and air-sea flux scheme with enthalpy coefficients provided a better forecast for ESCS Hudhud. The track errors on day 1 to day 4 was 61 km, 73 km, 85 km, and 96 km, respectively, and absolute error in terms of MSW was about 8 ms−1, 7 ms−1, 2 ms−1, and 8 ms−1 respectively at 9-km horizontal resolution. © 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH, AT part of Springer Nature.