Spatial scale dependency of errors in satellite cloud track winds

Abstract

Abstract Low level satellite cloud motion winds (CMWs) have been generated from METEOSAT data for Intensive Observing Periods (IOPs) during the FRONTS87 experiment of the Mesoscale Frontal Dynamics Project. CMWs were produced for a range of template sizes, ranging from 4∗4 infra-red pixels to 32∗32 infra-red pixels (note that in the area of interest an infra-red pixel is approximately 5∗10 km: in addition to templates comprising equal numbers of pixels on each side, winds were generated using templates comprising twice as many pixels in the E-W direction as in the N-S). During the IOPs the wind field was also observed using, among other facilities, dropsondes. The satellite winds were then compared with an objective analysis of the dropsonde data, which was averaged to the scale of the satellite winds in question. The relationship between template geometry (in infra-red pixels) and rms vector error (in m/s) is shown in Table 1. The large error for small template sizes occurs because noise in the radiances significantly affects the position of the peak in the correlation surface. The increase of error with template size at large template sizes occurs because the satellite wind is not representative of the wind field in the whole of the template. On the largest scale, satellite winds may represent the translation of synoptic scale weather systems rather than meteorological wind fields.