Hybrid AI-enhanced lightning flash prediction in the medium-range forecast horizon

Area of Science:

• Meteorology and Atmospheric Science
• Artificial Intelligence and Machine Learning
• Computational Science

Background:

• Traditional deterministic algorithms, based on physics and mathematical models, are foundational in scientific disciplines.
• Artificial intelligence (AI) offers powerful capabilities for pattern recognition in large datasets, complementing traditional approaches.
• Numerical Weather Prediction (NWP) models, like the European Centre for Medium-range Weather Forecasts (ECMWF), are crucial for forecasting.

Purpose of the Study:

• To develop an AI-enhanced algorithm for predicting lightning flash occurrence.
• To optimize the mapping of meteorological features from NWP models into lightning predictions.
• To compare the performance of the AI-enhanced algorithm against traditional deterministic methods.

Main Methods:

• Utilized meteorological features predicted two days ahead by the ECMWF NWP model.
• Developed an AI-based strategy to learn the mapping from meteorological features to lightning flash occurrence.
• Evaluated the prediction capability and performance metrics (Recall and Precision) of the AI-enhanced algorithm.

Main Results:

• The AI-enhanced algorithm demonstrated significantly higher prediction capability than the fully-deterministic ECMWF algorithm.
• Achieved a peak Recall of approximately 95% within a 0-24 hour forecast interval.
• Outperformed the ECMWF model's 85% Recall at the same Precision level.

Conclusions:

• AI integration substantially enhances lightning flash prediction accuracy.
• The developed AI-enhanced algorithm offers a superior alternative to traditional deterministic methods for lightning forecasting.
• This approach highlights the potential of combining NWP data with AI for improved severe weather prediction.