Livestock farms are recognized sources of ammonia emissions, impacting nearby regions’ fine dust particle concentrations, though the full extent of this impact remains uncertain. Air dispersion models, commonly employed to estimate particulate matter (PM) levels, are heavily reliant on data quality, resulting in varying levels of accuracy. This study compares the performance of both air dispersion models and spatiotemporal deep learning models in estimating PM concentrations in Republic of Korea’s livestock-farming areas. Hourly PM concentration data, alongside temperature, humidity, and air pressure, were collected from seven monitoring stations across the study area. Using a 200 m × 200 m prediction grid, forecasts were generated for both 1 h and 24 h intervals using the Graz Lagrangian model (GRAL) and a one-dimensional convolutional neural network combined with the long short-term memory algorithm (1DCNN-LSTM). Results highlight the potential of the deep learning model to enhance PM prediction, indicating its promise as an effective alternative or supplement to conventional air dispersion models, particularly in data-scarce areas such as those surrounding livestock farms. Gaining a comprehensive understanding and evaluating the advantages and disadvantages of each approach would offer valuable scientific insights for monitoring atmospheric pollution levels within a specific area.

Despite ample evidence that high levels of particulate matter (PM) are associated with increased emergency visits related to respiratory diseases, little has been understood about how prediction processes could be improved by incorporating real-time data from multipoint monitoring stations. While previous studies use traditional statistical models, this study explored the feasibility of deep learning algorithms to improve the accuracy of predicting daily emergency hospital visits by tracking their spatiotemporal association with PM concentrations. We compared the predictive accuracy of the models based on PM datasets collected between 1 December 2019 and 31 December 2021 from a single but more accurate air monitoring station in each district (Air Korea) and multiple but less accurate monitoring sites (Korea Testing & Research Institute; KTR) within Guro District in Seoul, South Korea. We used MLP (multilayer perceptron) to integrate PM data from multiple locations and then LSTM (long short-term memory) models to incorporate the intrinsic temporal PM trends into the learning process. The results reveal evidence that predictive accuracy is improved from 1.67 to 0.79 in RMSE when spatial variations of air pollutants from multi-point stations are incorporated in the algorithm as a 9-day time window. The findings suggest guidelines on how environmental and health policymakers can arrange limited resources for emergency care and design ambient air monitoring and prevention strategies.