Joint Centre for Mesoscale Meteorology (JCMM)
Z Wang, K A Browning and G A Kelly
An auto-adapting tracking technique, independent of the first guess to wind field, has been designed to extract cloud motion wind (CMW) vectors from geostationery satellite images. It has been implemented as a PC-based experimental cloud-motion wind-inferring system. The system uses procedures to improve the computing efficiency and increase the amount and reliability of cloud motion information. The resulting cloud motion wind vectors, each of which is marked with a quality flag (QF), are compared with model winds and direct observations obtained during the Fronts and Atlantic Storm Track Experiment (FASTEX).
The comparison of the CMW vectors with operational numerical model analyses (and 3-hour forecasts) from the United Kingdom Meteorological Office shows that the bias and rms speed difference, and the rms vector difference are -0.6, 3.2 and 6.1 ms-1 for a particular setting of quality flag (QF>1600). Comparison with the operational numerical model 3-hour forecasts from the European Centre for Medium-Range Weather Forecasts gives slightly worse results in general but it appears that CMW fields are closer to ECMWF model at low levels and to UKMO model at high and medium levels.
The comparison with direct observations from dropsondes, radiosondes, buoys and ships shows that, given QF>1600, the bias and rms speed difference and rms vector difference of the cloud motion wind vectors averaged over all levels are 0.4, 5.5 and 8.3 ms-1, respectively. These values are marginally better than the numerical analyses (and 3-hour forecasts) from either the UKMO model which gives corresponding values of -0.2, 5.7 and 9.5 ms-1, respectively, or from the ECMWF model which gives corresponding values of -0.5, 7.2 and 9.6 ms-1, respectively.
The low values of bias and rms speed difference
and rms vector difference of the cloud motion wind vectors are good indicators
that the method works well in the five cases.