Extended Abstract (1.6M)P1.3
Using a reflectivity “scale filter” to improve performance of the Storm Cell Identification and Tracking (SCIT) algorithm
Gregory J. Stumpf, CIMMS/Univ. of Oklahoma and NOAA/NSSL, Norman, OK; and W. D. Zittel, A. Witt, T. M. Smith, V. K. McCoy, and S. Dulin
The National Severe Storms Laboratory (NSSL) Storm Identification and Tracking Algorithm (SCIT) was implemented into the NEXRAD system and has been in operational use for about six years. During this time, it has become apparent that the algorithm faces shortcomings in its time association abilities. Most of these problems can be traced back into the horizontal and vertical association techniques in the algorithm, and are basically inherent in any algorithm which uses heuristic rules and centroid to centroid associations. Given approximately 100 storm events ranging from 10-60 minute lifetimes, and from 5 locations nationwide, the present time association technique only works about 73% of the time. In other words, for a given storm event, for one out of every four volume scans, the storm ID will be reassigned and all trends will be restarted. The NSSL partnered with the NEXRAD Radar Operations Center (ROC) to study various methods to improve SCIT time association. The first attempt to improve the algorithm included modifying some of the heuristic rules in the vertical association technique. This improved the time association success rate to 84%. However, a less costly alternative was subsequently tested which involved no changes to the SCIT source code. Instead, “scale filters” were applied to the initial reflectivity field designed to smooth out the smaller scale reflectivity features at near ranges to the radar while maintaining the larger scale features at all ranges. By applying the scale filter prior to the SCIT detection component, the time association success rate was improved to 96%.
Poster Session 1, 20th IIPS Poster Session (HALL 4AB)
Monday, 12 January 2004, 2:30 PM-4:00 PM, Hall 4AB
Previous paper Next paper