Um estudo numérico e estatístico da variabilidade do vento na Amazônia Oriental (Amapá e Pará - Brasil).

Abstract

The main objective of the study was to use the mesoscale model BRAMS (version 5.2) to simulate the wind variability, for the purpose of characterization and wind potential evaluation in the states of Amapá and Pará (Brazil) in 2009. The methodology was applied indexes and statistical refinement with Artificial Neural Networks (ANN) for validation and adjustments of the atmospheric model outputs. As a result, it was found that in Amapá average wind speed time was higher in Macapá with a value of 3,23 m/s and; Pará was in Soure with 2,61 m/s. Macapa (AP) and Soure (PA) had also the higher values in wind speed of the monthly scale with 3,23 and 3,00 m/s, respectively. After the ANN and use simulation to validation, noticed a statistically significant increase in numerical simulation, which increases the credibility of BRAMS to characterize the wind variability in different time scales. Moreover, the wind direction analysis, the mesoscale model was efficient representation in all seasons, including their magnitudes, except in Bethlehem to the predominant direction as was East and reproduced Northeast model. In the case studies for rainy and dry seasons, it was observed that both the BRAMS as ANN represented the wind speed variability efficiently in all seasons except Macapa that atmospheric model underestimated. Statistical indices (BIAS, RMSE and r) applied to the data comparisons were satisfactory for BRAMS and the ANN in which the Pearson coefficient of at most stations showed moderate correlations (0,40-0,69) reaching correlations very strong (0.90-1,0). Then, the PDF Weibull parameters indicated that BRAMS best simulated shape parameter in most points of study and the RNAs showed better representation of the scale parameter. Therefore, the use of atmospheric models to simulate the wind variability is an important tool, especially when there are no records of observational data in remote locations. And, in the case of Amazon, the low density of weather stations in the vast region derail detailed studies to monitor places with wind potential.