Uncertainty quantification and predictability of wind speed over the Iberian Peninsula. Issue 7 (12th April 2017)

• Record Type: Journal Article
• Authors: Fernández‐González, S.; Martín, M. L.; Merino, A.; Sánchez, J. L.; Valero, F.
• Is Part Of: Journal of geophysical research. Volume 122:Issue 7(2017)
• Page Start: 3877
• Page End: 3890
• Publication Date: 2017-04-12
• Subjects: ensemble prediction system -- ensemble spread -- wind -- uncertainty quantification -- predictability -- Iberian Peninsula; Atmospheric physics -- Periodicals; Geophysics -- Periodicals; 551.5
• DOI: 10.1002/2017JD026533
• Languages: English
• ISSNs: 2169-897X

Abstract

Abstract: During recent decades, the use of probabilistic forecasting methods has increased markedly. However, these predictions still need improvement in uncertainty quantification and predictability analysis. For this reason, the main aim of this paper is to develop tools for quantifying uncertainty and predictability of wind speed over the Iberian Peninsula. To achieve this goal, several spread indexes extracted from an ensemble prediction system are defined in this paper. Subsequently, these indexes were evaluated with the aim of selecting the most appropriate for the characterization of uncertainty associated to the forecasting. Selection is based on comparison of the average magnitude of ensemble spread (ES) and mean absolute percentage error (MAPE). MAPE is estimated by comparing the ensemble mean with wind speed values from different databases. Later, correlation between MAPE and ES was evaluated. Furthermore, probability distribution functions (PDFs) of spread indexes are analyzed to select the index with greater similarity to MAPE PDFs. Then, the spread index selected as optimal is used to carry out a spatiotemporal analysis of model uncertainty in wind forecasting. The results indicate that mountainous regions and the Mediterranean coast are characterized by strong uncertainty, and the spread increases more rapidly in areas affected by strong winds. Finally, a predictability index is proposed for obtaining a tool capable of providing information on whether the predictability is higher or lower than average. The applications developed may be useful in the forecasting of wind potential several days in advance, with substantial importance for estimating wind energy production. Key Points: Uncertainty of an ensemble prediction system was estimated Ensemble spread was used as an estimator of model predictability Several tools were developed to facilitate decision making