Uncertainty-Aware Graph Neural Reconstruction of Urban Temperature Fields from Sparse Sensors under Deployment Constraints
Researchers have developed a new graph neural network (GNN) framework designed to reconstruct urban temperature fields from sparse sensor data. This model not only predicts temperature but also quantifies predictive uncertainty, aiding in heat-risk analysis. The framework was evaluated using data from the Montreal area and demonstrated superior performance compared to traditional methods like inverse distance weighting and ordinary kriging, particularly under budget constraints for sensor placement. AI
IMPACT This research offers a novel approach to environmental data reconstruction, potentially improving urban planning and climate monitoring.