Toward Effective Source Apportionment Using Positive Matrix Factorization: Experiments with Simulated PM_2.5 Data.

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Chemical Engineering

Background:

• Particulate matter (PM2.5) impacts air quality and human health.
• Positive Matrix Factorization (PMF) is a common receptor model for source apportionment.
• Understanding PMF factor relationships to real emission sources is crucial for refining modeling strategies.

Purpose of the Study:

• To clarify the link between PMF-resolved factors and actual emission sources.
• To enhance the PMF modeling strategy for improved source identification.
• To evaluate the interpretability of PMF factors using various statistical metrics.

Main Methods:

• Speciated PM2.5 data from a chemical transport model were analyzed using PMF.
• Goodness-of-fit metrics (χ², R²) and root mean square difference analysis were employed.
• Factor interpretability was assessed with and without a priori knowledge and factor rotation.

Main Results:

• PMF factors often represent combined sources, not distinct ones.
• Optimal factor numbers should balance data explanation (R² > 0.95) without overfitting.
• Increased temporal data resolution aids minor source identification; factor rotation improves interpretability.

Conclusions:

• The number of factors and data resolution significantly impact PMF source apportionment accuracy.
• Incorporating prior knowledge and careful factor rotation enhance model interpretability.
• Uncertainty weighting and data variability can divert PMF solutions, necessitating careful validation.