A systematic examination of a random sampling strategy for source apportionment calculations

Area of Science:

• Environmental science
• Earth science
• Atmospheric science

Background:

• Estimating source contributions in environmental matrices is crucial across scientific disciplines.
• Traditional methods using linear equations for source apportionment can introduce statistical bias.
• This bias arises from assuming independence between calculated fractions and source distribution variability.

Purpose of the Study:

• To evaluate a random sampling (RS) strategy for addressing statistical bias in source apportionment.
• To assess the robustness, accuracy, reproducibility, and predictability of the RS methodology.
• To compare RS with numerical integration for two-source scenarios including source variability.

Main Methods:

• Development and application of a random sampling (RS) strategy.
• Utilizing rationally designed synthetic datasets for investigation.
• Comparison with a numerical integration solution for a two-source model.

Main Results:

• The random sampling (RS) methodology demonstrates robustness and accuracy.
• RS ensures reproducibility and predictability in source contribution estimation.
• Source profile variability impacts both precision and central tendency (mean/median) of contributions.

Conclusions:

• The random sampling (RS) strategy effectively accounts for statistical bias in source apportionment.
• This method provides a more reliable approach compared to traditional algebraic solutions.
• Acknowledging source variability is essential for accurate precision and contribution estimates.