Comparison of Objective Supercell Identification Techniques Using an Idealized Cloud Model

Abstract

Abstract The accuracy, reliability, and skill of several objective supercell identification methods are evaluated using 113 simulations from an idealized cloud model with 1-km horizontal grid spacing. Horizontal cross sections of vorticity and radar reflectivity at both mid- and low levels were analyzed for the presence of a supercell, every 5 min of simulation time, to develop a “truth” database. Supercells were identified using well-known characteristics such as hook echoes, inflow notches, bounded weak-echo regions (BWERs), and the presence of significant vertical vorticity. The three objective supercell identification techniques compared were the Pearson correlation (PC) using an analysis window centered on the midlevel storm updraft; a modified Pearson correlation (MPC), which calculates the PC at every point in the horizontal using a small 3 km × 3 km analysis window; and updraft helicity (UH). Results show that the UH method integrated from 2 to 5 km AGL, and using a threshold value of 180 m2 s−2, was equally as accurate as the MPC technique—averaged from 2 to 5 km AGL and using a minimum updraft threshold of 7 m s−1 with a detection threshold of 0.3—in discriminating between supercells and nonsupercells for 1-km horizontal grid spacing simulations. At courser resolutions, the UH technique performed best, while the MPC technique produced the largest threat scores for higher-resolution simulations. In addition, requiring that the supercell detection thresholds last at least 20 min reduced the number of false alarms.