Spatio–Temporal Environmental Monitoring Using IoT Sensor Networks with ConvLSTM and Graph Neural Networks

Abstract

The rapid expansion of smart cities has increased the demand for reliable environmental monitoring systems capable of detecting dynamic changes in urban climate and potential hazards such as pollution accumulation and fire outbreaks. Internet of Things (IoT) technologies enable continuous environmental sensing through distributed sensor networks. However, the complex spatial and temporal dependencies within environmental data pose significant challenges for traditional analytical models. This paper proposes a hybrid spatio-temporal deep learning framework that integrates Convolutional Long Short-Term Memory (ConvLSTM) networks with Graph Neural Networks (GNN) to analyze environmental data collected from IoT sensors. The ConvLSTM module captures temporal dependencies in multivariate sensor readings, while the GNN component models spatial relationships among sensor nodes. Experimental results demonstrate improved prediction accuracy compared with conventional models. The framework supports smart city applications including environmental monitoring, urban heat analysis, and early fire detection.