Weather Radar Data assimilation Impact in Very Short Range Forecasting - a Squall Line Case Study in Brazil

Due its wide extension,  Brazil  has several areas that are prone to produce severe storms (strong winds, heavy rainfall, lightning, flooding) which may lead to downbursts and tornadoes. Such phenomena produce wind speed that area greater than 30m/s and may cause damages on many high structures such as transmission lines and towers. One of the areas with frequent occurrence of severe storms is located in the  west of Paraná State, southern Brazil,  due to the influences of low level jets that bring humidity from tropical region, mainly. Paraná Meteorological System (SIMEPAR) operates a dual polarization S-Band weather radar in the region. Numerical weather prediction is essentially a boundary conditions problem which means that better initial conditions relate to better forecasts. The aim of this study was to improve model skill in short-range forecasts (about 1 hour) by assimilation of radar reflectivity, radial velocity along with automatic surface meteorological stations included as part of a project detailed in another paper in this conference (Calvetti et al., 2016).Since December 2015 a series of anemometers were installed in four energy towers within the weather radar range to study the behaviour of wind around and in these structures. Although the time series is too short and no extreme event was observed,  a case was chosen to validate the proposed methodology. On February 18th, 2016 at 18UTC  a gust of 27ms-1  was recorded in a sonic anemometer at 28m high associated by a convective-cell embedded in a squall line in that region.This study used the model  Weather Research Forecast  (WRF) with three nested grids from  9, 3 and 1km horizontal resolution centered on the radar site. The boundary conditions used were obtained from the model Global Forecast System (GFS) with a horizontal resolution of 0.25 degree latitude and longitude. Two sets of experiments were performed: first initializing the  model at 12UTC with four different sets of physics parameterizations, without data assimilation to identify the best configuration for that location; and the second set of experiments used the model output at 17UTC as first guess to the assimilation, which was over the 1km grid. Nested feedback carried out information throughout the domains.The model using Lin et al. microphysics (experiment M2) showed a better representation of the atmosphere state when compared to ETA microphysics (experiment M1) and  Goddard microphysics (experiment M3). On the other hand, changing PBL scheme to ACM2 in M2 (experiment M4) generated less convection when compared to observations and the other experiments.While the model without assimilation was not able to generate the convection as observed with radar data, the high-resolution simulation with radar data assimilation yield a development of the convection cells compared to the observations, although in some areas the reflectivity simulated were overestimated. The spatial distribution of reflectivity, and the magnitude of the wind forecasted by the model were similar to those observed, although the reflectivity amplitude were again overestimated in some areas. These preliminary results encourage further investigations in radar data assimilation for short-range forecast. Quality control is a major issue that should be investigated thouroughly, and in particular when regarding radial velocity. Using polarimetric variables can improve the quality control and is already in the process of analyis of this project. This work is part of PD-6491-0259/2012 - "Wind gust impact in falling towers events" project supported by COPEL generation and transmisstion S.A. (COPEL GeT) as part of the Research and Technological Development Program of Brasilian Eletric Energy Sector regulated by Nacional Agency of Eletric Energy (ANEEL).