Evaluating Microphysics, Cumulus, and Lightning Parameterization Schemes in WRF Model for Thunderstorm Simulation Over East India

Abstract

Lightning are the electrical discharges that result from non-inductive charging processes within thunderstorms, where collisions between different hydrometeors such as graupel, hail, and ice crystals leads to the separation and buildup of electric charges. Therefore, selection and tuning of parameterization schemes, particularly for Microphysics (MP), Cumulus (Cu), and Lightning (LP) processes, play a critical role in enhancing model performance and predictive accuracy.

This study evaluates the performance of the Weather Research and Forecasting (WRF) in simulating lightning for a severe thunderstorm event on 14 May 2022, over Eastern India (West Bengal and Jharkhand). The Indian Lightning Location Network (ILLN) recorded a peak 30-minute flash count of around 8000 flashes. A total of 57 combinations of MP-Cu-LP schemes were tested using a three nested domain (27 km, 9 km, and 3 km grid spacing) and analyzed the output of the innermost domain (3 km).

Results shows that the LP2 scheme (based on 20 dBZ reflectivity) significantly outperformed LP1 (updraft-based), particularly when combined with the Kain-Fritsch (KF) cumulus scheme enabled in the inner domain. Among MP options, Goddard, Morrison, and WDM-6 schemes produced the best results. The overall best performing scheme combinations: LP2-KFon-Goddard, LP2-KFon-Morrison, LP2-KFon-WDM5, and LP2-KFoff-WDM6 achieved high correlations (0.91–0.94) with observed lightning flash counts.

Vertical profile analyses revealed that the thermodynamic properties were consistently simulated across parameterization schemes, but microphysical differences influenced the lightning prediction accuracy. The best-performing combination (LP2 – KFon – MP 7 – Goddard) provided balanced cloud water and ice-phase dynamics, while the worst performing scheme (LP1–KFoff–Morrison) overestimated cloud water and ice content, leading to inaccurate lightning predictions. Therefore, the differences in model performance were not due to representation of large-scale thermodynamic or moisture transport processes but rather stem from representation of microphysical processes, particularly cloud water retention, precipitation efficiency, and ice-phase dynamics. Enhancing the representation of riming efficiency and hydrometeor phase transitions could enhance model accuracy in future lightning prediction