An Application of Deep Learning in the Zhuoshui River Basin for Multi-Station PM10 Forecast

Air pollution, particularly particulate matter (PM₁₀), presents a critical environmental and public health challenge, with the Zhuoshui River Basin in Taiwan being a severely affected region. One-day-ahead multi-station PM₁₀ forecasting is essential for effective air pollution management. In this study, we propose a deep learning architecture that integrates 3D image datasets and time series data, enabling the extraction of key information from heterogeneous inputs. The model significantly enhances forecasting accuracy compared to benchmarks, achieving R² improvements of 15–70% and RMSE reductions of 6–25%.

Regional PM₁₀ forecasting is crucial for protecting public health, as PM₁₀ exposure is linked to severe respiratory and cardiovascular risks and exacerbation of pre-existing conditions. Accurate forecasts enable authorities to issue timely warnings, implement mitigation measures, and allocate resources efficiently. Seasonal PM₁₀ forecasting is equally important, as air quality exhibits significant seasonal variations driven by meteorological and environmental factors. Our analysis reveals that the proposed model performs best during summer, achieving the smallest R² and largest RMSE improvements, while performance decreases in winter due to adverse conditions like temperature inversions and stagnant air masses.

These seasonal insights are critical for developing targeted strategies, such as stricter emission controls and public health advisories during winter months when PM₁₀ levels are highest. Moreover, accurate seasonal forecasts provide essential guidance for long-term urban and regional planning, including green infrastructure placement, enhancement of public transportation policies, and development of resilient air quality management systems. By integrating advanced deep learning models into air quality management frameworks, this research contributes to protecting public health and fostering sustainable development in the Zhuoshui River Basin.