Semester of Graduation

May 2022

Degree

Master of Science (MS)

Department

Geography and Anthropology

Document Type

Thesis

Abstract

Lightning strikes are incredibly common and potentially hazardous. Lightning can lead to loss of life, property damage, or can trigger other hazardous events like wildfires. They are also the result of incredibly complex physical processes in the atmosphere, making it challenging to predict strikes. However, given the extensive data and robust analytical techniques available, machine learning methods could be used to establish data relationships between lightning strikes and ground-level atmospheric measurements. Such a model could be used to provide any surface-based weather station with a method for predicting incoming strikes, making it easier to avoid the potentially hazardous effects of lightning. This analysis attempts to build a lightning prediction model using measurements collected during the month of June 2018 from stations in three major South Louisiana cities: Baton Rouge, Lake Charles, and New Orleans. Surface-level measurement of relative humidity, surface-level pressure, and wind speed from nearby weather stations are used in the model to relate to lightning activity. An additional model was developed using particulate matter to represent pollution in the air and the amount of cloud condensation nuclei available for storm formation. Data were aggregated at thirty-minute intervals, with machine learning models testing if they were able to predict a strike in the next thirty-minute intervals. These models were compared to a baseline model based on CAPE (Convective Available Potential Energy) and several of the models outperformed this baseline, specifically those with pollution represented. These results indicate that a model could exist that assists with the prediction of lightning. Further analysis would use similar techniques with additional variables such as maximum daily pollution level and seasonal information and a finer temporal resolution to predict on a finer timescale.

Committee Chair

Jill Trepanier

DOI

10.31390/gradschool_theses.5508

Available for download on Thursday, April 03, 2025

Share

COinS