Improved 4-DVar Simulation of Indian Ocean Tropical Cyclones Using a Regional Model

Abstract

The performance of the 4-D variational (4-DVar) data assimilation system over the 3-D variational assimilation system is investigated for the simulation of tropical cyclones (TCs), using the weather research and forecasting model. Two TCs (cyclone Thane and cyclone Hudhud) that formed over the North Indian Ocean have been considered in this paper. Surface and upper air observations from Global Telecommunications Systems in combination with ocean surface winds and satellite derived atmospheric motion vectors are assimilated cyclically at an interval of 6 h. The analysis is then integrated for a period of 72 h in the free forecast mode for both the assimilation runs. The 4-DVar run results for cyclone Thane shows significant impact on rainfall forecast, with an average equitable threat score (ETS) twice that of 3-DVar run, while the 4-DVar run for cyclone Hudhud shows a little improvement in ETS. The average initial error in mean sea level pressure for analysis fields for 4-DVar run is found to about half of the 3-DVar counterpart for the case of cyclone Hudhud. Improvement in 4-DVar simulation of maximum surface wind speed is found to be 7% and 14% for cyclone Thane and Hudhud, respectively. Also, the 4-DVar run for cyclone Thane resulted in an average reduction of 10% in track error simulation, while the same for cyclone Hudhud revealed a little improvement.