Quality control of global solar radiation data with satellite-based products

  1. Urraca, R. 2
  2. Gracia-Amillo, A.M. 3
  3. Huld, T. 3
  4. Martinez-de-Pison, F.J. 2
  5. Trentmann, J. 1
  6. Lindfors, A.V. 4
  7. Riihelä, A. 4
  8. Sanz-Garcia, A. 56
  1. 1 Deutscher Wetterdienst, Offenbach, Germany
  2. 2 Universidad de La Rioja
    info

    Universidad de La Rioja

    Logroño, España

    ROR https://ror.org/0553yr311

  3. 3 Joint Research Centre
    info

    Joint Research Centre

    Bruselas, Bélgica

  4. 4 Finnish Meteorological Institute
    info

    Finnish Meteorological Institute

    Helsinki, Finlandia

    ROR https://ror.org/05hppb561

  5. 5 Tokyo Women's Medical University
    info

    Tokyo Women's Medical University

    Tokio, Japón

    ROR https://ror.org/03kjjhe36

  6. 6 University of Helsinki
    info

    University of Helsinki

    Helsinki, Finlandia

    ROR https://ror.org/040af2s02

Mostrar afiliaciones +
Revista:
Solar Energy

ISSN: 0038-092X

Año de publicación: 2017

Volumen: 158

Páginas: 49-62

Tipo: Artículo

DOI: 10.1016/J.SOLENER.2017.09.032 SCOPUS: 2-s2.0-85029694007 WoS: WOS:000418974500007 GOOGLE SCHOLAR

Otras publicaciones en: Solar Energy

Resumen

Several quality control (QC) procedures are available to detect errors in ground records of solar radiation, mainly range tests, model comparison and graphical analysis, but most of them are ineffective in detecting common problems that generate errors within the physical and statistical acceptance ranges. Herein, we present a novel QC method to detect small deviations from the real irradiance profile. The proposed method compares ground records with estimates from three independent radiation products, mainly satellite-based datasets, and flags periods of consecutive days where the daily deviation of the three products differs from the historical values for that time of the year and region. The confidence intervals of historical values are obtained using robust statistics and errors are subsequently detected with a window function that goes along the whole time series. The method is supplemented with a graphical analysis tool to ease the detection of false alarms. The proposed QC was validated in a dataset of 313 ground stations. Faulty records were detected in 31 stations, even though the dataset had passed the Baseline Surface Radiation Network (BSRN) range tests. The graphical analysis tool facilitated the identification of the most likely causes of these errors, which were classified into operational errors (snow over the sensor, soiling, shading, time shifts, large errors) and equipment errors (miscalibration and sensor replacements), and it also eased the detection of false alarms (16 stations). These results prove that our QC method can overcome the limitations of existing QC tests by detecting common errors that create small deviations in the records and by providing a graphical analysis tool that facilitates and accelerates the inspection of flagged values. © 2017 The Author(s)