Data-driven Optimal Sensor Placement for Aerosol Monitoring in the UK
The research developed a system using a trained convolutional Gaussian neural process (ConvGNP) to find optimal locations for aerosol monitoring stations across the UK. The model predicts aerosol optical depth, and candidate sites are scored based on information gain metrics. This framework is computationally efficient and uses historical data to verify if the chosen uncertainty criteria match real-world predictive performance under different weather conditions.