Structure recognition in LOFAR lightning data
: An algorithmic approach to tracking lightning leaders

Student thesis: Master's Thesis

Abstract

Lightning has been a mysterious phenomenon for centuries. Amongst the biggest questions
in lightning physics today are the initiation and propagation of lightning leaders. In recent
years, lightning mapping technology has become a vital tool for lightning research. The LOFAR
telescope has proved an excellent tool for lightning mapping. To analyse the LOFAR data more
effectively, we need an automated system. This work represents the first steps towards this goal,
in which we design two algorithms that are able to recognize branches in a data set. The first one
is the light cone method, which uses the causal relationship between the events to reconstruct
the branches. The voxel algorithm on the other hand breaks the given data down into cubes and
uses their connections to group the data points into branches.

We compare their performance and find that the voxel algorithm is better in dealing with stray
points. However, the light cone algorithm is less likely to lose points during the reconstruction
phase. Both algorithms performed well on multiple subsets and their reconstructions allowed us
for the first time to systematically study different branch parameters. As such, we found a strong
indication for increased activity around branching points. In addition, we studied the opening
angle between branches after splitting and discovered the angle to always be between 70° and
110°. We also suggest several improvements to the light cone and voxel algorithm, as well as
multiple new methods.
Date of Award2021
Original languageEnglish
SupervisorStijn Buitink (Promotor) & Hershal Pandya (Co-promotor)

Keywords

  • Lightning
  • Structure recognition

Cite this

'