Prediction of daily PM(2.5) concentration in China using partial differential equations

Accurate reporting and forecasting of PM(2.5) concentration are important for improving public health. In this paper, we propose a partial differential equation (PDE) model, specially, a linear diffusive equation, to describe the spatial-temporal characteristics of PM(2.5) in order to make short-term prediction. We analyze the temporal and spatial patterns of a real dataset from China’s National Environmental Monitoring and validate the PDE-based model in terms of predicting the PM(2.5) concentration of the next day by the former days’ history data. Our experiment results show that the PDE model is able to characterize and predict the process of PM(2.5) transport. For example, for 300 continuous days of 2016, the average prediction accuracy of the PDE model over all city-regions is 93% or 83% based on different accuracy definitions. To our knowledge, this is the first attempt to use PDE-based model to study PM(2.5) prediction in both temporal and spatial dimensions.