Comparison of Two Algorithms for the Reduction of Noise Events in Orbital Optical Lightning Data

Abstract:

The Geostationary Lightning Mapper (GLM) will be launched as part of the GOES-R satellite. The GLM on-board software is designed to allow a large amount of non-lightning data (noise) to be transmitted to the ground. This is done so that the ground software can remove the noise while at the same time preserving the weak lightning events in the data stream. The ground software has a noise filter that removes events that do not meet certain coherency requirements. The filter utilizes the location, time, and amplitude of the event, along with the background level to determine the likelihood that the event is actually due to lightning and not noise. Due to various constraints, the technique does not use all relevant information to determine the likelihood that the event is actually due to lightning. A more complete filter that uses the full extent of the information available, including clustering results, should produce better results. However, the extra coding and complexity needed to implement the full clustering based filter may not justify the slightly better results. To test the various filter options, this study uses GLM proxy data generated from numerous ground based and orbital sources that mimic the expected characteristics of the GLM lightning event data. These proxy data sets have noise data added, again based on the expected characteristics of GLM noise data. The full proxy data sets (noise and lightning) are filtered by the two methods and the results compared.