• No se han encontrado resultados

3. EL DEBIDO PROCESO EN EL PROCEDIMIENTO CONTENCIOSO ADMINISTRATIVO

3.3. LA PRUEBA

The research context thoroughly discussed in the first section of this chapter was the inspiration which defined this work. The statistical models employed in the research of wind forecasting have proved to be advantageous for capturing the characteristics of wind speed. The major drawback though is that there is not converging evidence on which is the optimal method. This is due to the fact that these models depend on the time series and on the timescale of the

⁵This is valid up to the time this study is submitted and the statement refers to papers that have been published only in English.

forecasts. Consequently, although they perform satisfactorily in each individual study, providing better understanding of wind time series forecasting, the picture is yet far from complete. Similar conclusions are drawn from the use of the same models specifically in the long-term topic.

Data Collection and Analysis

This chapter contains identical parts from three papers [99] (published journal) [118] (published

conference) [119] (published journal)

T

he main thrust of this chapter is to provide information about the data used in this study.The needs of the research called for drawing on data from various sources. Such sources consisted of historic long-term onshore measurements from several Met stations of the UK Met Office across the country as well as reanalysis data. The succeeding procedure of setting criteria for including/excluding particular stations is amply discussed, while details about factors that possibly affect wind speed measurements are provided. Information then is presented on designing the structure of a database which was a critical step for the conduct of the research. Overall, this chapter is organised in sections based on the chronological phases carried through for the purposes of collection, filtering, analysis, and organisation of data.

3.1 Sources of Data

3.1.1 Surface Observations from the UK MIDAS via the BADC

The UK Met Office produces a data set of land surface observations from 1853 to the present date which is freely available for research purposes. This argument as it stands alone was sufficient for engaging with the Met Office in order to get hold of the data. The vast amount of historic onshore records in combination with the well spread coverage over the UK constituted the drive for using primarily that source. Another reason that limited the present research to the use of

data provided by the Met Office was a combination of the hypothesis aimed to be tested with the particular family of statistical models which was chosen. As mentioned in the literature review, the autoregressive models require more data in order to capture the characteristics of time series. At the same time getting hold of data for a concurrent period that derive from tall masts (e.g. above 60 m agl) was not feasible as there are a few meteorological masts in the UK that combine the desired height (i.e. hub height) and the long period of measurements that was necessary in the study. An alternative solution would be to engage developers that have a large fleet of 10 m masts across the country but the drawback under this scenario would be that these masts would not record data for the same period. Also, by engaging developers or owners of wind farms, delays would be added in the progress of the research. This would be inevitable as non disclosure agreements would need to be signed off before any data were sent for analysis. For these reasons it was decided that a flexible, fast and reliable solution that was not compromised in terms of both spatial and temporal coverage would be the data produced by the UK Met Office. This data set is held on the Met Office Integrated Data Archive System (MIDAS) and is available via the British Atmospheric Data Centre (BADC) [120]. The MIDAS data set contains a large number of observations covering a variety of meteorological parameters including mean wind speed [121]. Wind observations are typically 10-minute means in knots (kt) and are typically made at a height of 10 m above the land surface. However, this has not necessarily been the case historically as will be discussed later in this chapter. Since this research has two different objectives the data used for each case differs slightly.

• The part that deals with the variability of wind speed and will be presented in Chapter 4 uses two different groups of data. Although both are retrieved from MIDAS, their distinct difference lies in the number of the stations contributing and the period of the series. One group used records for the period 1983 - 2011 and the second for the period 1957 - 2011 (see Chapter 4 for details). In order for the readers to be able to identify which dataset or which stations each time this study is referring to, these will be known henceforth as BADC-57 and BADC-7 respectively.

• Similarly to above, the part that deals with the long-term forecasting, and will be presented in Chapter 5, uses data from surface stations. In order to accomplish homogeneity between the different stations in terms of time, the year 1957 was set as the starting year for all stations used to generate the long-term forecasts. The specific year was set because, after applying several criteria for the stations that would be included in

the analysis (see section 3.2 below), the earliest common year for all the stations was found to be 1957. Hence, this part uses also the BADC-7 dataset.

3.1.2 ERA-40 Reanalysis Dataset

The hypothesis set to be tested in this study aimed to investigate whether other atmospheric variables except wind, can increase the efficacy in the long-term forecasts. Reanalysis data comprise a useful source that is rich in both spatial and temporal coverage. Most importantly, reanalysis datasets incorporate a vast amount of observations that, through an assimilation system, offer useful insights related to climate. Similarly to the onshore stations, the present study required data that were dated back to the starting date of the onshore measurements. One of the reanalysis data that covered a period longer than 40 years is the ERA-40 which is also freely available for someone to download. Thus, it was decided for the purpose of the study to use the ERA-40 dataset. The ERA-40 reanalysis was produced by assimilating a large number of different meteorological datasets, including satellite measurements, ship-borne and buoy observations, land-based surface observations, upper air measurements and remote sensing observations [122]. The assimilated data have been output onto both a 2.5◦× 2.5◦ grid and a 1◦ × 1◦ grid at six hourly intervals covering the period 1957-2002. It should be noted that land-based surface wind speed observations were not used as input to the assimilation, though the assimilating model produces output surface wind speeds on the regular array of grid points, including those over land. In this work, 10 m values of the uand v′wind components from the 1◦×1◦grid were extracted. uand v′are the zonal and meridional wind components. For vector fields, such as wind velocity, the zonal component refers to eastwards wind while the meridional component refers to northwards wind. Then, the magnitude of wind speed, u, was calculated using Pythagoras theorem:

u =

Documento similar