Flow Analysis of Cloud Images from Geostationary Satellites

In this paper, motion analysis of cloud images from a geostationary satellite has been examined by the method of optical flow. The original model of the optical flow is adjusted to cater for the compressible property of clouds. Also, the spiral movement equation is added to derive the movement of occluded depressions which circulate around a vortex. The analysis aims to support rainfall analysis and prediction. Geostationary satellites are a valuable source of rainfall information due to the availability of a global view of clouds at an acceptable spatial and temporal resolution. However to retrieve the information from the satellite images is a significant challenge. For example, precipitation peaks while the cloud area is rapidly growing and reduces at the time of maximum cloud area [4], Visible (VIS) and Infrared (IR) channels of the satellites can see only the top-of-the-clouds, not rain at the surface of the earth. Moreover, how a cloud changes with time reflects atmospheric instabilities that occur and most instabilities lead to precipitation. As a consequence, we need some descriptions of cloud motion and pattern changes as an explicit link to rain rate. To derive the velocity of an object in three dimensional space from a sequence of two dimensional images or optical flow, Horn and Schunck [2] introduced the fluid dynamics constraint reducing the ambiguity of the velocity field, thus making the recovery of an object’s motion possible. The idea is also relevant to cloud motion, which is a special case of fluid motion and consists of very complex motion dynamics [5]. Furthermore, its probability distribution allows representation of the uncertainties in the optical flow computation [3].