University Of Tehran Press Pollution 2383-451X 11 3 2025 07 01 Forecasting the Short-Term Changes of Surface Ozone and NO2 during a Festival Event Using Stochastic and Neural Network Models 576 592 100989 10.22059/poll.2025.377011.2387 EN Ebin Antony Department of Information Technology, Kannur University, Kannur, India Keerthi Lakshmi Department of Physics, Sree Krishna College Guruvayur, Kerala, India- 680102 Sunil Kumar Department of Information Technology, Kannur University, Kannur, India Nishanth Theeyancheri Department of Physics, Sree Krishna College Guruvayur, Kerala, India- 680102 0000-0001-7117-3572 Jalaja Kunnath Department of Applied Science and Humanities, Nehru College of Engineering and Research Centre, Pampady, Thrissur, Kerala-680588 Satheesh Kumar Department of Atomic and Molecular Physics, Manipal Academy of Higher Education, Karnataka, India- 576104 Annie Sabitha Paul Department of Applied Sciences, Govt. Engineering College Kannur, Kerala India-670563 Journal Article 2024 05 25 Air pollution is one of the most destructive environmental issues on the local, regional, and global level. Its negative influences go far beyond ecosystems and the economy, harming human health and environmental sustainability. By these facts, efficient and accurate modelling and forecasting the concentration of air pollutants are vital. Hence, this work explores investigate the time series components of surface ozone (O3) and its precursor nitrogen dioxide (NO2) and develops a model for predicting O3 variations produced by intense fireworks during the Vishu festival over Kannur. Time series methods using Stochastic and Recurrent Neural Network (RNN) and Seasonal Autoregressive Integrated Moving Average (SARIMA) models are considered the most accurate tools for estimating air pollution trends due to their logical flexibility. Model performance is evaluated based on statistical measurements indicating an increasing trend in O3 concentration of 0.11 ppb/year and NO2 of 0.18 ppb/year. Based on the analysis, we found that the SARIMA model shows better accuracy with a Mean Squared Error (MSE) of 0.55 and a Root Mean Squared Error (RMSE) of 0.74. The broader implications of this study highlight the applicability of advanced time series forecasting techniques for air quality monitoring during short-term pollution events. https://jpoll.ut.ac.ir/article_100989_ab175312fb52e6fab464c1143393fa10.pdf