Use of mobile phone traffic data for the assessment of air pollution exposure of urban population.

Abstract

Environmental epidemiological studies use air pollutants concentrations to relate exposure with living population for estimation of health effects. This relationship is normally inferred by matching concentrations with addresses of residence of population. However, a large part of population spent most of the time away from its living place, introducing an incorrect assessment of exposure and of related health effects. Consequently, mobility of population should be considered for a proper assessment of exposure. Mobile phone traffic data (TIM Big Data Challenge 2015) were collected for more than a million of persons living/working in Rome, Italy, during March and April 2015 and used to derive 15 minutes time resolved population maps over an irregular grid with a minimum resolution of 0.26x0.34 Km2. Two datasets were available providing the number of persons as total and gender/age/postal code separated respectively. The latter data were matched in time and space with air pollution model results to produce time resolved cumulative population exposures during the studied period. Strong variability of exposures has been detected as an effect of both mobility and time/spatial changing in pollutants concentrations. Quartiles values are found to be up to ±50 and ±30% wider of their median values for NO2 and PM2.5 respectively. A comparison with the stationary approach based on census data allowed detecting the inability of latter in estimating the actual variability of population exposure. Significant underestimations of the amount population exposed to daily WHO guideline level for PM2.5 has been detected, with 25% and 80% on April 2015 for the census and the mobile phone based methods respectively. Very small differences (up to a few ug/m3) on exposure are detected for gender, age ranges and postal codes selected population.