Using the WRF Model to Refine NCEP CFSv2 Reanalysis Atmospheric Profile: A Southern Brazil Test Case

ABSTRACT. The vertical distribution of atmospheric parameters is pivotal to understand atmospheric dynamic processes. This study analyzes the use of the Weather Research and Forecasting (WRF) model to generate atmospheric profiles as an alternative for a local radiosonde. We performed simulations in the WRF model for two dates with distinct weather conditions (01/09/2018 – cloudy and 03/14/2018 – clear-sky). The simulated data were compared with radiosondes launched in the study area, both (i) along the entire profile and (ii) with focus on the Planetary Boundary Layer (PBL) – for the variables water vapor mixing ratio ( q ), potential temperature (?) and wind speed ( Speed ). The results showed a high correlation between the simulated and observed profiles, with most R values higher than 0.9. Bias and RMSE, respectively, ranged between -1.29 – 0.66 g/kg and 0.48 – 2.01 g/kg for  q ; -0.52 – 0.25 K and 0.44 – 3.00 K for ? and -0.20 – 1.31 m/s and 1.61 – 2.77 m/s for  Speed . The difference between the profiles from different WRF nested grids was minimal, suggesting that a resolution of approximately 12 km is enough to a good trade-off between detailing and computational cost. WRF model seems suitable to simulate fine atmospheric profiles. Keywords:  vertical profiles; numerical weather prediction; meteorology; remote sensing; planetary boundary layer. Uso do Modelo WRF para refinar perfis atmosfericos de reanalise NCEP CFSv2: um estudo de caso no Sul do Brasil RESUMO. A distribuicao vertical dos parâmetros atmosfericos e fundamental para entender os processos dinâmicos da atmosfera. Este estudo analisa o uso do modelo  Weather Research and Forecasting  (WRF) na geracao de perfis atmosfericos, como uma alternativa a radiossondagens locais. Foram realizadas simulacoes com o WRF para duas datas em condicoes de tempo distintas (09/01/2018 – alta nebulosidade e 14/03/2018 – ceu claro). Os dados simulados foram comparados com radiossondagens lancadas na area de estudo, (i) ao longo de todo o perfil e (ii) com foco na camada limite planetaria (PBL) – para as variaveis razao de mistura do vapor de agua ( q ), temperatura potencial (?) e velocidade do vento ( Speed ). Os resultados mostraram uma alta correlacao entre os perfis simulados e observados, com a maioria dos valores de R superiores a 0,9. Vies e RMSE, respectivamente, variaram entre -1,29 – 0,66 g/kg e 0,48 – 2,01 g/kg para  q ; -0,52 – 0,25 K e 0,44 – 3,00 K para ? e -0,20 – 1,31 m/s e 1,61 – 2,77 m/s para  Speed . Minimas diferencas entre os perfis das diferentes grades aninhadas sugerem que uma resolucao horizontal de aproximadamente 12 km e um bom balanco entre detalhamento e custo computacional. O WRF mostrou potencial na simulacao de perfis atmosfericos refinados. Palavras-chave : perfis verticais; previsao numerica do tempo; meteorologia; sensoriamento remoto; camada limite planetaria.