Development of Pattern Recognition Algorithms to Detect Intense Convective Storms from Multispectral Satellite Imagery

This work presents a new method to detect overshooting top (OT) clouds, which are produced by intense convective updrafts and are often associated with hazardous weather. The method uses visible and infrared (IR, $11\ \mu \mathrm{m}$) input images from any low-earth-orbiting (LEO) or geostationary (GEO) satellite imager. The described IR algorithm improves upon those previously developed by minimizing use of fixed criteria and incorporating pattern recognition analyses. The visible detection branch of the algorithm uses two-dimensional (2D) Fast Fourier Transform (FFT) to detect “cauliflower” texture of the visual appearance of OTs as well as the gravity waves emanating away from intense updrafts. The OT detection product is validated using a diverse sample of human identified OT occurrences and has shown a false detection rate of 18%, which drops to 1% when the visible detection is coupled the IR-based probability product.