Pattern recognition filtering and bidimensional FFT-based detection of storms in meteorological radar images

Important climate changes are today observed around the world, frequently yielding destructing precipitations. To identify and follow the evolution of storms in real-time, meteorological radar images collected in Setif (Algeria), Bordeaux (France), and Dakar (Senegal), are processed. Template and pattern recognition-based filters are firstly used to remove the ground clutter and keep the precipitation echoes unchanged. Bidimensional FFT is then applied to the filtered images, showing that the Fourier spectra characterising convective clouds differ significantly from those of stratified ones. This difference can be usefully employed by the radar operators to quickly detect the formation of violent storms. To forecast the growing of rainfall clouds and their motion, the related radar echoes are reconstituted using inverse FFT. It is found that the 26 first harmonics are sufficient to both rapidly and accurately reconstitute the surface of clouds whereas 82 distinct harmonics are needed to well reproduce their reflectivity.