that provide wind speed distributions. These may be adequate for micrositing of turbines, because the output of wind plants is likely to be less sensitive to small errors in the distribution of wind direction than wind speed.
Hub-Height Shear and Natural Turbulence
For turbine selection and load estimation, it is also important to know the expected distribution of shear and natural (non-wake) turbulence. As turbine sizes have grown, this requirement has become more critical. Current sources of this information are surface measurements, which need to be scaled up to hub height, and numerical atmospheric models. Neither of these methods is adequate. Surface measurements are widely scattered. Moreover, the techniques for scaling wind from surface measurements generally involve either a power law approach or a logarithmic wind profile. These methods are commonly applied without regard to aerodynamic roughness of the sea surface, atmospheric thermodynamic stability, or the effect of rapidly changing atmospheric or oceanic conditions, leading to large errors in these approaches. There is no accurate method for scaling surface measurements to make accurate hub-height turbulence measurements. The situation is not much more encouraging with numerical models. Such models generate estimates both of wind shear and turbulence, but there is little evidence that validates their performance for hub-height shear and turbulence. Extreme wind shears and other supplemental data from hurricanes are available from archives of global positioning system (GPS) dropsonde wind profiles collected from the National Oceanic, Atmospheric, and National Oceanic and Atmospheric Administration (NOAA) and Air Force “Hurricane Hunter” aircraft since 1997.
Air Temperature and Atmospheric Surface Pressure
Air temperature is needed in conjunction with atmospheric pressure primarily to calculate the distribution of air density at prospective wind plant sites. Unlike dynamic variables, this information is well known both from surface measurements and from numerical models.