3.3 Polarimetric Tornado Detection With an Adaptive Neuo-fuzzy Inference System

Wednesday, 9 January 2013: 11:00 AM
Room 18A (Austin Convention Center)
Yadong Wang, CIMMS/Univ. of Oklahoma, Norman, OK; and T. Yu

The upgrade of Weather Surveillance Radar-1988 Doppler (WSR-88D) into dual-polarization model is ongoing. Tornado debris signatures (TDS) defined as high reflectivity, anomalously low cross-correlation coefficient, and negative or close to zero differential reflectivity have been observed by polarimetric radars at different frequencies. In the current work, an Adaptive Neuro-fuzzy Inference System (ANFIS) was developed, which has the capability of integrating tornado shear, spectral, and debris signatures. There are several advantages of using the ANFIS: 1.) ANFIS framework is flexible to integrate partial or all the tornado signatures; 2.) ANFIS can still provide robust tornado detection results if one or more input variables are not associated with obvious tornado signatures; and 3.) parameters implemented within the ANFIS can be optimized through a self-learning procedure. The developed ANFIS was evaluated with 15 tornado events with scales from EF0 to EF4 and ranges from 16 km to 133 km. The archived level II data was acquired from the NCDC website. The total output probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) are 86%, 13%, and 76.2%, respectively.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner