Imputação de dados horários de velocidade do vento no território continental do Brasil

Abstract

Wind is an important meteorological variable for both rural and urban economic planning, such as energy production. However, there is a shortage of Nationwide studies aiming statistical database provision and evaluation on the time scale, which is partially due to the lack of flawless series of observed data across the Brazilian mainland. Thus, the objective of this research is to fulfill a database of wind speed, having time sampling without failures for the entire continental area of Brazil. Initially, the dataset was retrieved from 449 automatic surface stations, a number then reduced to 421, against which the multiple input method has been executed. Data have time sampling and coverage of the period between January 1st, 2010 and December 31st, 2020. Both the original time series and the inputs present a significant amout of outliers, which affect the quantitative analysis. A total of 5,560,709 values have been input and, after that, values of standard deviation and variance have been lower, with differences of 0.06 and 0.19, respectively, indicating a decrease of the data variability. Highest values have been observed in Southern and Southeastern regions. Lowest time values have been observed in the Northeastern region. Some years (2010, 2016, 2020 and 2021) present an average value increase. September presented the highest value variation of data before and after the input. Data input presented a higher value in the first quarter, Q1, in most of the months, when compared with the Q1 of the data not input. In the temporal analysis, there was a large variation between the original data and the input data between 11:00 AM e 08:00 PM. Yet, the outlier of highest level occurred at 09:00 AM. The study reveals that differences between descriptive statistics of incomplete series and the series after inputting have not been dissonant, which indicates that data input have been generated as values near the median, with no impact to the original series characteristics.