On the Improved Predictive Skill of WRF Model With Regional 4DVar Initialization: A Study With North Indian Ocean Tropical Cyclones

The predictive skill of the weather research and forecasting model is examined, and the improved performance with the 4-D variational (4DVar) data assimilation scheme over the 3-D variational (3DVar) scheme is quantified for the simulation of four tropical cyclones over the Indian region by generating a large number of analysis/forecast samples. Satellite radiance, satellite-derived winds, and conventional observations are assimilated cyclically for both 3DVar and 4DVar experiments at an interval of 6 h. Each of the analyses is then subjected to a short-range free forecast, lasting 48 h. The analysis fields are found to capture the features of all the cyclones more realistically when the model is initialized with the 4DVar assimilation. Free forecasts from the 4DVar analysis fields are superior in simulating the intensity and position of the storm for most of the times. Rainfall simulation also shows marked improvement in terms of the equitable threat score for the 4DVar run. Furthermore, in the case of cyclone Phailin, it is noted that the 4DVar run could successfully capture the rapid intensification phase. On an average, the simulation of intensity and position has shown an improvement of 2%-43% and 22%-57%, respectively, by the 4DVar run at different forecast lead times.