A Short-Term Cloud Forecast Scheme Using Cross Correlations

Abstract

This paper describes a cloud forecast technique using lag cross correlations. Cloud motion vectors are retrieved at a subset of points through multiple applications of a cross-correlation analysis. An area in the first of two sequential frames of satellite data is correlated with surrounding areas in the second frame to find the one surrounding area best correlated. The location difference of the areas defines the displacement vector. An objective analysis is used to define displacements at every satellite pixel throughout the domain and smooth the local inconsistencies. Using these displacements, forecasts are then produced with a backward trajectory technique. This scheme was tested using two IR satellite images of the same scene a half-hour apart and found to generate realistic, high-quality forecast IR pixel images. Results demonstrate improvements over persistence and movable persistence for forecasts of a few hours? length. The technique is visually appealing, since forecasts are created in pixel images of the same form and resolution as the initializing satellite data, permitting animation. It is also computationally inexpensive.