A novel algorithm for convective cell identification and tracking based on Optical Character Recognition Neural Network

Abstract

Identifying the core of the convective cloud is paramount important to the identification of the Mesoscale Convective System (MCS) and its evolution is of great significance in the weather and climate system. To study the initiation and development of convective core systems, an automatic tracking algorithm named “Convective cell Identification and TRAcking (CITRA)” has been developed to identify and track the convective cells in MCSs using Doppler Weather Radar reflectivity images based Optical Character Recognition using Long Shortterm Memory Neural Network. Further, CITRA algorithm calculates the convective cell physical properties such as centroid, convective core size and area, duration, maximum extent, distance and direction from the Radar centre. CITRA algorithm tested on 1255 convective cells identification and tracked 90 distinct convective system families along with physical properties through their evolution during the monsoon periods of 2017–2019. The details of CITRA algorithm is described in the present paper.

KEYWORDS:

• Mesoscale Convective System
• convective cell
• identification
• tracking
• Optical Character Recognition
• DWR reflectivity images

Acknowledgements

The DWR images are downloaded from the MOSDAC website. The authors are very greatly thankful to the Anaconda Inc. and OpenCV community for supporting the project through their valuable open-sourced content. The authors are greatly thankful to the anonymous reviewers for their evaluation and valuable suggestion to improve the paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).