Quantifying sources of uncertainty in reanalysis derived wind speed. (August 2016)

• Record Type: Journal Article
• Authors: Rose, Stephen; Apt, Jay
• Is Part Of: Renewable energy. Volume 94(2016)
• Page Start: 157
• Page End: 165
• Publication Date: 2016-08
• Subjects: CFS reanalysis -- Wind integration -- Linear mixed-effect model; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2016.03.028
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: Reanalysis data are attractive for wind-power studies because they can offer wind speed data for large areas and long time periods and in locations where historical data are not available. However, reanalysis-predicted wind speeds can have significant uncertainties and biases relative to measured wind speeds. In this work we develop a model of the bias and uncertainty of CFS reanalysis wind speed than can be used to correct the data and identify sources of error. We find the CFS reanalysis data underestimate wind speeds at high elevations, at high measurement heights, and in unstable atmospheric conditions. For example, at a site with an elevation of 500 m and hub height of 80 m, a CFS reanalysis wind speed of 8 m/s is 0.2 m/s higher to 1.3 m/s lower than the measured wind speed. We also find a seasonal bias that correlates with surface roughness length used by the reanalysis model during the spring season. The corrections we propose reduce the average bias of reanalysis wind speed extrapolated to hub height to nearly zero, an improvement of 0.3–0.9 m/s. These corrections also reduce the RMS error by 0.1–0.4 m/s, a small improvement compared to the uncorrected RMS errors of 1.5–2.4 m/s. Highlights: We develop a model to correct for bias and uncertainty in CFS reanalysis wind speed. Altitude, height above ground, and atmospheric instability are sources of bias. Under-prediction increases with altitude and hub height but decreases with wind speed. The model we develop reduces reanalysis bias 61–99% and RMS error by 1–31%.