<p>The study investigates the impact of assimilation of both conventional and non-conventional observations on improving the initial condition and forecast of super cyclonic storm (Su-CS) Kyarr developed over the Arabian Sea using the WRF-3DVAR technique. The results show that using 3DVAR data assimilation significantly improves the accuracy of the analysis. The initial condition was enhanced using the 3DVAR system, and the simulated intensity and track of the cyclone were verified with the available IMD best-fit track dataset. Results indicate that the track and intensity of Su-CS Kyarr were influenced by the improved initial time. The mean track error was about 298&#xa0;km and 198&#xa0;km in CNTL (control run) and DA data assimilation, respectively. The track errors from day 1 to day 7 were improved by 12.6%, 45.5%, 41.2%, 54.7%, 55%, 46%, and 1.7%, respectively, in the DA experiment compared to CNTL. The predicted intensity of the storm in terms of the MSW and MCP was slightly better predicted in the DA experiment due to the improved initial condition. The accumulated rainfall from 00 UTC on October 25 to 00 UTC on November 2 was analyzed using the experiment 2400 observed RMSE of approximately 97.5 and a higher CC of about 0.7 with the observation, with a GPM estimate. Moreover, DA experiment showed a remarkable performance with high POD, PAG values and low FAR values than CNTL. The storm structure in terms of temperature anomaly, reflectivity, and horizontal wind was well enhanced by the DA experiment. In addition, simulations conducted using cold start and cyclic mode assimilation for Su-CS Kyarr suggests that cold start mode is more beneficial for capturing the cyclone track and intensity. The experiment cold start mode that showed the best performance was tested on a larger number of cases to evaluate the model performance.</p>