Data assimilation for tropical cyclone-induced rainfall forecasting for central Viet Nam

Tóm tắt

Data assimilation plays a particularly important role in numerical weather prediction models (NWPs). It ingeste observational data into the initial fields of NWPs, improving their initial conditions to better represent the actual atmospheric conditions. This enhancement consequently leads to improved forecast accuracy of NWP. Globally, three-dimensional variational assimilation (3DVAR) and four-dimensional variational assimilation (4DVAR) methods have been widely used, with 4DVAR considered the most advanced technique. This study will be focused on data assimilation methods using the WRFDA system with WRF-ARW Core. This arrtice was investigated cases without data assimilation of 3DVAR and 4DVAR with different assimilation windows. Evaluations the initial fields demonstrate that both 3DVAR and 4DVAR methods effectively improve the model's initial fields by assimilating observational data. This is evidenced by the fact that the RMSE  of the 3DVAR and 4DVAR analysis fields is consistently smaller than the RMSE of the original model initial fields when compared to observations and for all fields, including wind, temperature, moisture, and pressure. This study further reveals that the 4DVAR method offers superior improvements in the initial fields compared to the 3DVAR method. Under the same  conditions, the RMSE of 4DVAR analysis fields is generally smaller than that of 3DVAR analysis fields. The evaluation based on the POD and FAR indices indicates that the forecast skill of the 4DVAR case is better than that of 3DVAR, especially at larger rainfall thresholds above 50 mm, which is clearly reflected in the 24-hour accumulated rainfall forecast.