Improving cloud-top divergence signals with a bilateral filter

Abstract

Severe weather intensity trends can be monitored from satellite imagery over regions with sparse radar coverage using novel products available from optical flow retrievals. For example, cloud-top divergence rendered from the retrieved brightness motions, provides an indirect measure of updraft intensity with time. Recent demonstrations have now shown that dense optical flow can render sub-storm scale (< 5 km) motions, which appear noisy to operational forecasters during warning operations evaluations. The bilateral filter, a spatial signal smoothing filter that retains large-scale features while attenuating signal noise, is an approach for removing any unwanted cloud-top divergence noise signals while preserving the large-scale signals forecasters use in practice. Little is currently understood, however, on how such filters modify observed trends and magnitudes in cloud-top divergence, in particular how such magnitudes change during and in advance of severe weather observations at the ground. Two bilateral filter sizes were evaluated on a mid-latitude supercell in the Great Plains and tropical convection off the Northeast coast of South America, and it was determined that a bilateral filter with a gaussian sigma size of 2.5 adequately removed unwanted signals for both cases while maintaining the 5 min severe weather lead time associated with cloud-top divergence. Additionally, filtered signals were reduced by ~38% to ~60% for sigma sizes of 2.5 and 5.0 respectively.