Thunderstorm Predictions Using Artificial Neural Network with Radiosonde Indices in Saumlaki Area

Thunderstorms are hydro-meteorological disasters that need to be vigilant. In response, various studies have been conducted to determine suitable prediction methods for forecasting thunderstorms, one of which involves using upper air data. However, the use of atmospheric stability threshold values as a reference may not apply universally to all locations due to differences in the characteristics of each region. In meteorology, especially for predicting specific events, machine learning methods such as artificial neural networks are widely used. Therefore, this study employs an artificial neural network approach to predict thunderstorms in the Saumlaki region based on radiosonde index data. The research compares the performance of machine learning models using the Pattern Recognition Neural Network (PNN) technique in predicting thunderstorms, specifically between models with a single hidden layer and double hidden layers with a certain number of neurons in each hidden layer, as previously used in other studies. Through this comparison, the study aims to provide insights into how accurately the PNN model can predict thunderstorms in the Saumlaki region.