A Contour-Based Algorithm for Automated Detection of Overshooting Tops Using Satellite Infrared Imagery

Abstract

Overshooting top (OT) is a proxy for deep convection with intense updrafts. Using Himawari-8 infrared window (IRW) and water vapor (WV) imagery, this paper presents a contour-based OT detection algorithm based on the premise that the OT center is in a cold region enclosed by the local coldest (innermost) brightness temperature (BT) contour. Relative to existing IRW-texture algorithms, the contour-based algorithm offers improvement in two aspects. First, the new algorithm utilizes the WV minus IRW BT difference to capture the deep-convection pixels, and then removes the warmer pixels far from the cold center. Thus, compared with directly using a simple 215-K BT threshold to determine the possible OT pixels, many false alarms are avoided. Second, the new algorithm utilizes the BT contours to roughly depict and separate the anvil clouds and OT regions, and the anvil cloud pixels used to estimate the mean anvil cloud BT are outside the OT regions. Therefore, the computed BT gradient is more credible compared with the simple 225-K BT threshold. The accuracies of the OT detection algorithms are evaluated with 159 Global Precipitation Measurement-observed OT events collected from July 2015 to July 2017. The results indicate that the contour-based algorithm performs better than the original IRW-texture algorithm. The contour-based algorithm achieves a probability of detection (POD) of 67% and a false-alarm ratio (FAR) of 46%. Compared to the original IRW-texture algorithm, the contour-based algorithm increases the POD by 9% and decreases the FAR by 42%.