Spatial imputation for air pollutants data sets via low rank matrix completion algorithm.

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Incomplete hourly air pollutant concentration data is a significant challenge in urban air quality monitoring.
• Existing imputation methods often struggle with large data gaps and spatial correlations.

Purpose of the Study:

• To propose and evaluate a novel spatial interpolation method for imputing missing air pollutant data.
• To assess the performance of Low Rank Matrix Completion (LRMC) against established techniques.

Main Methods:

• Utilized Low Rank Matrix Completion (LRMC), a spatial interpolation technique leveraging data correlation.
• Evaluated LRMC for imputing time series of NOₓ, O₃, SO₂, PM₂.₅, and PM₁₀ concentrations.
• Compared LRMC with nearest neighboring, mean substitution, regression, spline, spectral, and EM algorithms.

Main Results:

• LRMC demonstrated superior performance, especially with higher missing data ratios (e.g., 30%) at central monitoring stations.
• The method consistently produced robust imputations and was insensitive to the length of missing data gaps.
• LRMC achieved high goodness of fit (R²) for air pollutant time series imputation.

Conclusions:

• LRMC is a highly effective and robust method for imputing missing air pollutant concentration data.
• The technique successfully captures inherent spatial and temporal patterns in air quality data.
• LRMC offers a promising solution for enhancing urban air quality monitoring data completeness.