Estudo da velocidade do vento através de downscaling dinâmico em alta resolução sobre terreno complexo no Nordeste do Brasil

Abstract

Using advanced tools in wind flow modeling based on Numerical Weather Prediction (NWP) is essential for wind projects, since these techniques help to get a depth knowledge about the wind pattern within any geographical area. Dynamical downscaling is widely used in mesoscale models: a grid nesting method used to perform atmospheric simulations in high resolution at a low computational cost. In this work, we run numerical simulation using the mesoscale model Weather Research and Forecasting (WRF) for a nesting process of three grids in order to produce a high-resolution simulation in terrain area complex. For a performance test, we use a set of observed wind speed and air temperature data through August 2005 and it was obtained from an anemometric tower (50 meters) located in Belo Jardim/PE. In the validation methodology, statistical metrics such as the root mean square error, the standard deviation and Pearson's correlation were calculated between the observed and simulated datasets. Results show that using different grid nesting configurations significantly interferes the performance of the mesoscale model in representing the phenomena in the study region. Observational mean hourly data and the grid 01 showed an RMSE between 1.2 to 1.4 mps and a standard deviation around 0.97 to 1.9 mps. Grid 02 had a RMSE between 0.85 to 1.9 mps and a standard deviation ranging 0.85 to 1.3 mps. Grid 03 got a RMSE approximately 0.9 to 1.6 mps and a standard deviation of 0.75 to 1.1 mps. The lowest RMSE for all spacing grids was found at 4 pm local time. Overall, the wind speed diurnal cycle of grid 01 performed better during the first hours of the day and at night. This domain dimension can influence the results performance, since it inserts more information about the adjacent regions, predominating over the simulation of local winds. In the central hours of the day, the simulated wind speed was overestimated due to a higher estimate of the turbulence from the model, and during the night, it was underestimated. It is likely that the vertical temperature profiles may be more difficult to represent by the model due to its more stratified nature and, therefore, the turbulent flows estimated at this time of day.