Lightning Forecasting in Pre-Monsoon Season Using Non-Linear Autoregressive Artificial Neural Network

Abstract

Lightning is one of the most beautiful and dangerous phenomena in nature. It is an interesting and undetermined area of research in which information is vaguely defined. It has excellent potential to produce severe damage to living bodies and properties. The process of lightning is generally dependent on different meteorological parameters. The main objective of this study is to apply the concept of a nonlinear autoregressive network with exogenous inputs to an artificial neural network model (NARX-ANN) and predict the afternoon lightning in the pre-monsoon season. For this purpose, three meteorological parameters, namely atmospheric temperature (AT), relative humidity (RH), and stability parameter (z/L), have been taken as inputs to the proposed model. The performance of the model was evaluated on pre-monsoon data with prediction accuracy of 96.14%. Furthermore, results obtained from the seven skill scores have been evaluated, where False Alarm Rate (FAR) and Miss Rate (MR) were found near to zero. The result shows that the NARX-ANN model has minimum prediction errors and can be considered a suitable method for forecasting lightning.