Improving long-term air pollution estimates with incomplete data: A method-fusion approach

Abstract

Mobile air pollution monitoring is an effective means of collecting spatially and temporally diverse air pollution samples. These observations are often used to predict long-term air pollution concentrations using temporal adjustments based on the time-series of a fixed location monitor. Temporal adjustments are required because the time-series is often incomplete at each spatial location. We describe a method-fusion temporal adjustment that has been demonstrated to improve the accuracy of long-term estimates from incomplete time-series data. Our adjustment approach combines the techniques of using a log transformation to modify the air pollution samples to a near normal distribution and incorporates the long-term median of a reference monitor to mediate the effects of estimate inflation created by outliers in the data. We demonstrate the approach with hourly Nitrogen Dioxide observations from Paris, France in 2016.Method-Fusion Benefits:•Log transformations control for estimate inflation created by log normally distributed data.•Adjusting data with the long-term median, rather than the mean, controls for estimate inflation.•Produces more accurate long-term estimates than other adjustments independent of the pollutant being estimated.