Fig. 1: Comparative performance of TransNet and AGATNet across 13 forecast lead times for PM2.5 concentration predictions. | npj Clean Air

Fig. 1: Comparative performance of TransNet and AGATNet across 13 forecast lead times for PM2.5 concentration predictions.

From: TransNet: a transport-informed graph neural network for forecasting PM2.5 concentrations across South Korea

Fig. 1: Comparative performance of TransNet and AGATNet across 13 forecast lead times for PM2.5 concentration predictions.

Performance metrics comparing AGATNet and TransNet forecasts of surface PM2.5 concentrations at 170 monitoring stations across South Korea for the year 2021. a shows Mean Absolute Error (MAE; μg/m3), panel (b) shows Root Mean Square Error (RMSE; μg/m3), (c) shows Index of Agreement (IOA), and (d) shows Pearson correlation coefficient (R). Red circles connected by solid lines represent AGATNet performance, while blue squares connected by solid lines represent TransNet performance.

Back to article page