In the paradigm of Big Data, we have large data sets that have information that you want to discover. It is here where it enters the first stage of a comprehensive science of data. This discipline seeks to extract knowledge from large volumes of unstructured data, using converging estadistico-matematicas techniques from different disciplines. It is a continuation of the areas of data mining and predictive analytics combined, allowing to acquire knowledge about the behavior of different variables.
Figure 1. Some of the fields that influence the process of science of data
In this particular case, was to perform a detailed analysis of a series of data which include particularly the speed of winds and gusts with their respective addresses for one season of the core area of the pampas, Argentina. Front of a horizon of work in which more data are incorporated each time and limitations found in the mechanistic approaches, posed approach enables interesting information of a practical and simple way.
To begin with, proceeded to carry out an approach that we will call "naive", consisting of a first contact with the data, trying to illustrate graphically to extract knowledge from their behavior. After an instance which consisted in the evaluation of these null values, (in order to know whether they corresponded to the nature of the problem or were measurement errors), the figure was built 2 shows the values reported for 2014-2015 both wind speed and gust, color-coded according to the direction of them. While the intense colours (blue and green) represent winds and blasts that took place in North-South direction, those gray show which originated in the East-West direction.
Figure 2. Historical series of data for station of the pampas for speeds of wind (A) and (B) with its corresponding address blasts. (C) direction of wind (scale) and blast (color)
All the drawings in Figure 2 show a trend in which much of the winds in the North-South direction are of greater intensity than those that occur with East-West direction. These patterns are reinforced when the graph is analyzed (C) of Figure 2, where the winds and gusts addresses are compared and observed that although most of the cases these are coincidental, there are times of the year with marked Crosswinds (purple dots in yellow zone).
Figure 3. Relationship between speed and wind blasts according to the different times of the day
On the other hand, in Figure 3 you can see the variation in the relationship between the speed of the wind and gusts to the different moments of the day. Interacting with the chart you can select the time whose correlation is want know and appreciate the results. Of that figure we can also deduce that a strong correlation exist at all hours of the day between the wind speed and gusts.
Figure 4. Box-plot average wind speed for the different studied months
If we want to know the monthly seasonality we see Figure 4. There is that month-to-month variations occur in the values corresponding to the speed of the wind and a few outliers, arising only in July 2014 and in January 2015.
For greater detail on the study of the seasonality we turn to figure 5 in which we have an interactive chart to see the time seasonal variability for certain months of the year.
Figure 5. Box-plot mean wind speed for a day, according to the referred month
CAs we can see from the above figures, it is possible to obtain a great deal of information about the behavior of the wind.The applied analysis can be used in other stations in the region in order to build maps that allow us to understand the spatial development of the meteorological phenomena that dominate many agricultural practices, such as the application of fertilizers or agrochemicals. At BerecoLabs, we believe that an approach based on Science of data allows to understand the nature of the problem comprehensively, meeting the challenge of a large number of variables or information, allowing the extraction of knowledge that allows to develop solutions to the problems that affect the territories becoming intelligent.