A Survey of Methods and Architectures for Environmental Weather Monitoring and Prediction System Using IoT and Multi-Model Progressive Dense Self-Attention

Abstract

Environmental weather monitoring and prediction systems have gained significant importance due to climate change, increasing natural disasters, and the need for precise agricultural planning. Traditional forecasting methods often suffer from low spatial resolution and delayed predictions. The integration of Internet of Things (IoT), machine learning (ML), and deep learning (DL) techniques has revolutionized weather prediction by enabling real-time data acquisition and intelligent forecasting. Recent advancements, particularly multi-model progressive dense self-attention architectures, have further enhanced prediction accuracy by capturing complex spatio-temporal dependencies in environmental data. This survey presents a comprehensive review of methods and architectures used in IoT-based weather monitoring and prediction systems between 2020 and 2023. It focuses on sensor-based data acquisition, cloud/fog computing frameworks, and AI-driven predictive models such as LSTM, CNN, and attention-based networks. The study also highlights challenges such as data heterogeneity, energy efficiency, scalability, and model generalization. A comparative analysis is conducted based on accuracy, computational complexity, scalability, and real-time performance. The findings indicate that hybrid deep learning models combined with IoT frameworks outperform traditional statistical methods by providing localized and high-resolution predictions. Finally, the paper outlines future research directions emphasizing edge intelligence, federated learning, and attention-based architectures for next-generation weather forecasting systems.