Abstract
Operational weather prediction systems do not currently make full use of infra-red satellite observations that are affected by the presence of cloud. Observations that are affected by cloud are routinely discarded during pre-processing. This is because cloud causes large, unpredictable, and nonlinear changes in the observed radiances, and obscures the atmosphere underneath from view. This disrupts the finely-balanced calculations used to convert small changes in observed radiance into temperature and humidity profiles of the atmosphere. Areas that contain cloud are likely to be meteorologically interesting, so where information on the state of the atmosphere is most desired, it is also in shortest supply. This thesis explores the possibility of using the large changes over time of cloud-affected infra-red satellite observations to calculate the vertical component of wind. In order to explore the mathematical and practical issues of assimilating data from cloudy radiances, a study has been performed using an idealised single column atmospheric model developed for this purpose. The model simulates cloud development in an atmosphere with vertical motion and the effects on simulated infra-red satellite observations. An empirical method and a variational data assimilation system have been developed to process sequences of observations over a six hour time with the goal of calculating vertical velocity. These two methods combined allow vertical velocity to be determined with an RMS error of approximately 0.8 cm/s in 80% of cases. The system is capable of detecting the remaining cases where there is insufficient information in the observations to constrain vertical velocity. This result is the first step in the long term goal of using cloud-affected satellite imagery more effectively in operational weather prediction systems. The ability to use these observations in this way would improve the forecasting of severe weather events, helping to protect lives and property from loss or damage.