Theses and Dissertations

Advisor

Mercer, Andrew

Committee Member

Rudzin, Johna

Committee Member

Brown, Mike

Date of Degree

5-15-2026

Original embargo terms

Visible MSU Only 1 year

Document Type

Graduate Thesis - Campus Access Only

Major

Geoscience (Professional Meteorology/Climatology)

Degree Name

Master of Science (M.S.)

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

Lightning forecasting during the warm season in Mississippi remains a constant challenge in numerical weather prediction (NWP) due to the highly convective nature of thunderstorms and processes governing storm initiation, growth, and electrification. Warm season convection is often driven by mesoscale and boundary layer processes that are difficult for models to resolve. This study investigates warm season lightning predictability integrating Weather Research and Forecasting (WRF) model output with a graph neural network (GNN) based post processing framework. WRF simulations are conducted for June 1st-30th during 2001 2007 using North American Regional Reanalysis initial and boundary conditions, with Vaisala National Lightning Detection Network observations used for verification. Spatiotemporal predictors, derived from WRF, are ingested by a GNN that explicitly models spatial and temporal dependencies to produce probabilistic lightning forecasts. Forecast skill is evaluated using Brier scores and grid point contingency table statistics, with WRF derived probabilities serving as a baseline for comparison.

Download

Share

COinS