Identifying regional service function from PM_2.5 mass concentration throughout a city with non-negative tensor...

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Air pollution, particularly fine particulate matter (PM), poses significant health risks.
• Understanding spatial and temporal pollution patterns is crucial for effective environmental management.
• Beijing's ambient monitoring data provides a valuable resource for pollution analysis.

Purpose of the Study:

• To holistically examine pollution patterns in Beijing using time, day, and region dimensions.
• To introduce and apply the non-negative tensor factorization (NTF) method for pollution pattern identification.
• To correlate identified pollution patterns with area service functions and land use.

Main Methods:

• Utilized public fine particle concentration data from 35 ambient monitoring stations in Beijing.
• Applied a data-driven non-negative tensor factorization (NTF) method to distinguish pollution patterns.
• Calculated reconstruction correlation of tensors across sites, time, and days to evaluate model performance.

Main Results:

• Identified and annotated five distinct pollution patterns: traffic, industrial, residential, commercial, and steady.
• Each pattern exhibited unique temporal and daily characteristics.
• Achieved high model evaluation values with reconstruction correlation approaching 0.95-0.96.

Conclusions:

• The NTF method effectively identifies pollution patterns and their relation to area functions, outperforming traditional land-use classifications.
• The identified patterns offer insights into pollution sources and their dynamics.
• Findings support targeted public travel and control measures to improve Beijing's air quality.