Recent Advances in Environmental Weather Monitoring and Prediction System Using IoT and Multi-Model Progressive Dense Self-Attention: A Systematic Review

Abstract

Environmental weather monitoring and prediction systems are essential for applications such as agriculture, disaster management, and smart city development. Traditional numerical weather prediction methods, although reliable, often struggle to capture fine-grained spatial and temporal patterns and require significant computational resources. The integration of Internet of Things (IoT) and Artificial Intelligence (AI) has enhanced this domain by enabling real-time data collection and advanced predictive modelling. IoT sensors continuously gather environmental parameters such as temperature, humidity, rainfall, and wind speed, generating large-scale time-series data. Deep learning models, including Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, have been widely used for weather prediction, but they face limitations in capturing long-range dependencies. Transformer-based architectures and self-attention mechanisms address this issue by effectively modelling global spatial and temporal relationships, leading to improved forecasting accuracy. Hybrid models that combine CNNs, transformers, and graph-based techniques further enhance prediction performance and robustness. Despite these advancements, challenges such as high computational complexity, data heterogeneity, and scalability persist. Overall, multi-model hybrid architectures with dense self-attention mechanisms represent a promising approach for developing accurate, efficient, and scalable weather prediction systems.