S17
Comparison of RAP Forecast Wind Data with LIDAR Measurements in the Maryland Wind Energy Area

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Sunday, 4 January 2015
Daniel Wesloh, University of Maryland Baltimore County, Baltimore, MD; and S. Rabenhorst and R. Delgado

Wind interacts with or drives most meteorological phenomena, whether steering synoptic systems along their path, feeding dry air into thunderstorms, or driving the rain sideways in a downpour. The accurate determination of the speed and direction of the wind is therefore central to any effort to simulate the atmosphere or predict the weather. In order to check the National Weather Service's 13-km Rapid Refresh model (RAP), lidar wind measurements from the ocean east of Maryland during July and August 2013 were compared to the RAP 0-hour analyses and 3-hour forecasts for the same times at the same locations in one of the first comparisons of operational model output to offshore wind profiles. The forecast wind speeds were slower than the observed values, by 1.5 m/s for the 0-hour forecast and 0.8 m/s for the 3-hour forecast, with standard deviations relative to the lidar observations of 2 m/s and 3 m/s, respectively, at a height of 140 m. The 0-hour forecasts showed no clear pattern of changing error or bias with increasing height, while the 3-hour forecasts produced values that become both smaller and more variable relative to the lidar observations with increasing height. The wind directions given by the model showed a counterclockwise bias, 4° for the 0-hour forecast and 11° for the 3-hour, with model-relative standard deviations of 33° and 45°, respectively, at 140 m above sea level. The bias in each decreases slightly with height, but the variability relative to the observations remains the same over the range of observation heights. The change in bias as the forecast works its way forward from the analysis state to the 3-hour forecast reflects the changing nature of the model state as it evolves from the initialization state, fresh from data assimilation, to a forecast with the state fully under the sway of the model equations and starting to drift from reality. Proper assessments of wind resources in an area require a thorough understanding of these issues.