Leveraging NMF to Investigate Air Quality in Central Taiwan

Abstract

This study investigates air pollution in central Taiwan, focusing on key pollutants, including SO$_2$, NO$_2$, PM$_{10}$, and PM$_{2.5}$. We use non-negative matrix factorization (NMF) to reduce data dimensionality, followed by wind direction analysis and speed to trace pollution sources. Our findings indicate that PM$_{2.5}$ and NO$_2$ levels are primarily influenced by local sources, while SO$_2$ levels are more affected by transboundary factors. For PM$_{10}$, contributions from domestic and transboundary sources are nearly equal.

Figures (29)

• Figure 1: Central Taiwan Terrain Map: Stars indicate Industrial Air Quality Monitoring Stations; squares indicate Background Air Quality Monitoring Stations; and circles indicate General Air Quality Monitoring Stations.
• Figure 2: Visualization for hourly average concentration
• Figure 3: Visualization for monthly average concentration.
• Figure 4: The flow chart for proposed data analysis
• Figure 5: NMF for cophenetic correlation