72-hour real-time forecasting of ambient PM2.5 by hybrid graph deep neural network with aggregated neighborhood spatiotemporal information

•Hybrid graph neural network can realize 72 h real-time reliable prediction.•AOD data can provide extra constraint information for the prediction model.•Quantifying domain spatial information can reduce redundant information.•Expanding the modeling area in different directions is necessary for 72 h prediction.•The GNN_LSTM model can dynamic forecast the “source”-receptor relationship.