Relationships among daily mean and maximum wind speeds, with application to data quality assurance

• Published in: International journal of climatology 2006-01, Vol.26 (1), p.29-43
• Main Author: Graybeal, Daniel Y.
• Format: Article
• Language: English
• Subjects: daily maxima; data quality; Earth, ocean, space; Exact sciences and technology; External geophysics; extremes; Meteorology; peak‐gust factor; wind speed; Winds and their effects
• ISSN: 0899-8418; 1097-0088
• DOI: 10.1002/joc.1237

A growing number of climate change and variability studies, as well as applied research toward improving engineering design climatographies, require high‐quality, long‐term, extreme‐value climate data sets for accurate and reliable estimates and assessments. As part of a historical weather data rescue project of the US government, new data quality control procedures are being developed and applied for daily maximum wind speeds. Not only are existing quality assurance procedures mostly lacking for such data but the climatological relationships upon which such quality checks may be based are also grossly underexploited. Therefore, this study seeks to elucidate relationships among peak‐gust, fastest‐mile, and fastest 5‐min wind speeds, utilizing the peak gust factor model but generalizing it for these and other extreme wind‐speed elements. The relationship between peak‐gust factor and daily mean wind speed is also adapted for quality assurance and for a wider range of climates than previously studied. Fastest‐interval wind‐speed factors are found to follow Gaussian, gamma, or Weibull probability distributions, included within mixed models to handle zeros. Resistant prediction interval estimates about a resistant regression were developed for quality assurance of peak‐gust factor, given the daily mean wind speed. Flagging thresholds were estimated using parametric bootstrapping. Flag rates from 0.05 to 0.5% are in line with rates reported in the literature, from work with similar data sets; overall Type I and Type II error rates are in the range 0.03–0.3%. The approach outlined lends itself straightforwardly to application in data quality assurance. Copyright © 2005 Royal Meteorological Society.