Error characteristics of satellite-derived winds

Errors of winds derived from geostationary satellite measurements are estimated by analysis of differences between satellite and rawin observations. It is estimated that the NESS winds contain root mean square errors (rmse) of 4.7 m s-1 for low levels and 8.5 m s-1 for upper levels. These errors stem from the "noise" inherent in measuring cloud displacements and errors introduced by meteorological influences. "System noise," evaluated by three independent techniques, amounts to rms vector magnitudes of 2.5 to 3.0 m s-1 for picture pair vectors and about 4 m s-1 for manual vectors. These figures pertain only to clouds tracked with infrared imagery. The accuracy of these data can be increased by reducing system noise, but improving their meteorological value can be realized only by modifying procedures to enable clouds to be tracked in meteorologically active regions.