Investigating Lightning Parameterizations for Thunderstorm Forecasting Over Odisha

Abstract

In eastern India, thunderstorms and frequent lightning during the pre-monsoon season (March-May) result in significant human fatalities. Due to limited research on thunderstorm prediction, this study attempts to evaluate lightning flash counts using various lightning parameterization schemes in the Weather Research and Forecasting (WRF) model over Odisha, an eastern Indian state. Additionally, offline diagnostic methods are assessed for estimating lightning flashes from model-generated storm parameters. The simulated total lightning flashes are validated using observations from the Lightning Detection Network (LDN) for four selected events. To count the flash rates, the Price and Rind flash rate equations are applied using two methods; one based on Cloud Base Height (H) [Scheme 1] and other one based on the Maximum Vertical Wind Speed (W_max) [Scheme 2]. The results suggest that the Scheme 1 overestimates flash counts but has high probability of detection, whereas Scheme 2 underestimates flash counts but produces fewer false alarms. The spatial distribution of lightning flashes aligns well with accumulated rainfall, maximum reflectivity, and time evolutions patterns. As per the categorical performance diagram, Scheme 1 performs better in terms of detection probability, while both the schemes exhibit similar success ratios. Lightning Potential Index (LPI) is an indirect measure of lightning activity and cannot be evaluated using the previously mentioned performance metrics. However, LPI effectively identifies peak cloud-to-ground (CG) flash rates. Scheme 2 underestimates CG lightning making it less suitable for forecasting while Scheme 1 captures the spatial spread of CG flashes but requires additional filtering to reduce false alarms.