An initial condition that closely represents the true atmospheric state can minimize errors that propagate into the future, and could theoretically lead to improvements in the forecast. 3D-VAR allows us to combine concurrent three-dimensional observations into the numerical model to optimize their strengths and at the same time constrain their weaknesses. This study aims to evaluate and understand the impacts of 3D-VAR on the state of the art Weather Research and Forecasting (WRF) model that has a two nested domain setup. Our observation data are provided by NCEP which includes Surface (METAR, SHIPS, etc.) and Upper-Air (RAOB, ACARS, etc.) data. Previous month's forecasts are used to calculate the Background Error Statistics (BES) via the National Meteorological Center (NMC) me...[ Read more ]

An initial condition that closely represents the true atmospheric state can minimize errors that propagate into the future, and could theoretically lead to improvements in the forecast. 3D-VAR allows us to combine concurrent three-dimensional observations into the numerical model to optimize their strengths and at the same time constrain their weaknesses. This study aims to evaluate and understand the impacts of 3D-VAR on the state of the art Weather Research and Forecasting (WRF) model that has a two nested domain setup. Our observation data are provided by NCEP which includes Surface (METAR, SHIPS, etc.) and Upper-Air (RAOB, ACARS, etc.) data. Previous month's forecasts are used to calculate the Background Error Statistics (BES) via the National Meteorological Center (NMC) method via the gen_be utility provided by the WRF Data Assimilation (WRFDA) system. The domain configuration of our model covers China with an emphasis on Guangdong province, and has a two nested domains configuration with a resolution of 27km, 9km, and 3km. The improvements of the forecasts for all the scenarios are systematically compared and are quantified in terms of 2m temperature, 10m wind speed, sea level pressure and 2m relative humidity. We initialized a 4 days forecast (including 24 hours of spin-up time) every 24 hours for the month of June (summer case) and December (winter case). Results show that 3D-VAR provides significant improvements in correcting the BIAS and the RMSE for the winter case and post 24 hours forecasts show that improvements persist and have a growing tendency. The improvements contributed by 3DVAR in winter case is mainly associated with the 2m temperature and 2m relative humidity. On the other hand, applying 3DVAR on the summer case does not provides similar outcomes. The results show that the 3DVAR contributes little to no observable improvements for this period of time. Comparisons between these two periods were done to further understand the reason behind the distinct differences.