Skip to main content

Hierarchical Data Fusion Framework for Air Pollution Nowcasting

Start year: 2024

Summary: Development of a hierarchical data fusion framework for air pollution nowcasting. The research program is expected to address (i) Accuracy: updating beliefs from Dempster-Shafer rules as a generalisation of Bayesian inference while avoiding the use of Monte Carlo approximation, (ii) Reliability: LWSN data are more reliable by using dependable schemes triggered by DSI to select the most valid among the collocated sensor motes while reducing missing information and prolonging the sensor service life, and (iii) Scalability: the methodology can be extended to select the hyperparameters in a combination of various sources for the whole pipeline towards multiscale forecasting.