Objective detection of stripe patterns in satellite imagery caused by gravity waves: Lessons learnt from the southern hemisphere

Abstract

An algorithm to objectively identify regions of parallel stripes, which is based on Gabor filters plus subsequently applied grating cell operators, is reviewed and enhanced. The intended application is a meteorological one, to detect the presence of gravity waves which eventually may lead to atmospheric turbulence – a hazard to be dealt with in the daily routine work of aviation meteorologists. The waves are searched for in 7.3µm water vapour channel images recorded by the imager on board the Meteosat Second Generation geostationary satellites. Over the wide ocean areas of the southern hemisphere, there is ample opportunity for marine stratocumulus clouds to develop. They have the potential to form periodic patterns causing (undesired) responses from the grating detection algorithm. The proposed remedy is to exploit the much higher anisotropicity of gravity waves by constraining oneself to the signals associated with the prevailing texture direction. The pursued real-world application – with stripes undergoing deformation by the wind or modification by other atmospheric or orographic mechanisms – also reveals that one should not only search for regularity in the direction perpendicular to the examined texture direction, as done previously; rather, a certain range of search angles should be scanned to detect the regularities reliably. It is demonstrated how much the separability between gravity waves and unwanted signals improves as a result of those measures.