Development of PM2.5 forecasting system in Seoul, South Korea using chemical transport modeling and ConvLSTM-DNN  

<p>Ambient exposure to PM2.5 can adversely affect public health, and forecasting PM2.5 is essential for implementing protection measures in advance. Current PM2.5 forecasting systems are primarily based on the chemical transport model of Community Multiscale Air Quality (CMAQ) modeling systems and the Weather Research and Forecasting (WRF) model. However, the forecasting accuracies of these models are substantially constrained by uncertainties in the input data of anthropogenic emissions and meteorological fields, as well as inherent limitations in the models. </p><p>The PM2.5 concentrations in South Korea largely depend on the long-range transport (LRT) from China and local emissions; therefore, input feature data and machine learning algorithms were designed to reflect the spatial relationship between the source and receptor regions as well as the local temporal variation in PM2.5 in forecast regions. </p><p>The PM2.5 forecasting system developed in this study aimed at overcoming the limitations of CMAQ predictions as well as reflecting the spatial and temporal characteristics effectively by utilizing advanced hybrid machine learning algorithms of convLSTM and DNN. The proposed system was developed using forecast data from CMAQ and WRF, as well as observed PM2.5 concentrations and meteorological variables at monitoring stations in China and South Korea. It was then applied to PM2.5 forecasting in Seoul, South Korea. </p><p>The forecasting system was tested for 2015-2020 and evaluated for 2021. The performance of the hybrid model could effectively reduce the bias in the CMAQ forecast by improving the accuracy as well as the probability of PM2.5 episode detection.  </p><p> </p>