Real-Time Estimation of Population Exposure to PM2.5 Using Mobile- and Station-Based Big Data

Бин Чэн, Yimeng Song, Tingting Jiang, Ziyue Chen, Bo Huang, Bing XuInternational Journal of Environmental Research and Public Health 2018 · DOI

TL;DR

This paper introduces a novel method for real-time estimation of population exposure to PM2.5 in China by integrating mobile phone location data with ground-based air quality monitoring. This approach overcomes limitations of traditional census-based methods by dynamically accounting for human mobility and population distribution, offering a more accurate and granular assessment of exposure at a 3-hour interval.

What Problem Does This Paper Address?

Estimating population exposure to air pollutants like PM2.5 is crucial for understanding health risks. However, existing methods often rely on static census data, assuming constant population distribution and air quality throughout the day. This fails to capture the dynamic nature of human mobility and the varying levels of pollution people encounter as they move. Consequently, these static assessments can lead to significant biases, especially for short-term exposure impacts and cumulative health effects. The paper aims to bridge this gap by developing a method that incorporates real-time population dynamics and spatiotemporal variations in PM2.5 concentrations.

Methods

The core of the proposed method lies in integrating two key data sources:

  • Mobile Phone Locating-Request (MPL) Big Data: This data, sourced from Tencent, provides real-time location information from smartphone users. It's aggregated into 3-hour intervals and used to create pixel-based population density maps at a ~1.2 km resolution. This MPL data is then used to downscale static census data, creating dynamic population distribution maps.
  • Station-Based PM2.5 Observations: Hourly PM2.5 measurements from 1465 ground monitoring stations across China are used. These are combined with meteorological data (temperature, wind speed, visibility) and processed using a Geographically Weighted Regression (GWR) model to estimate PM2.5 concentrations across the entire study area at a ~1.2 km resolution.

The paper then combines these dynamic population distribution maps with the estimated PM2.5 concentrations using a population-weighted metric to estimate real-time population exposure. Furthermore, it extends this to estimate cumulative inhaled PM2.5 masses by incorporating human respiratory volume.

Datasets

  • PM2.5 Monitoring Station Data: Hourly ground-station PM2.5 measurements from 1465 stations across China (March 1-31, 2016).
  • Meteorological Data: 3-hourly measurements (temperature, surface wind speed, horizontal visibility) from 411 stations in China and 128 stations in buffer zones (March 1-31, 2016).
  • Mobile Phone Locating-Request (MPL) Big Data: Real-time location data from Tencent, aggregated to 3-hour intervals at a ~1.2 km resolution (March 1-31, 2016).
  • Population Census Data: Latest city-level population census data for China in 2014.

Key Insights and Ideas

  • Human Mobility is Key: The study strongly emphasizes that static population data is insufficient for accurate exposure assessment. Incorporating dynamic population distribution derived from mobile phone data significantly improves the realism of exposure estimates.
  • Spatiotemporal Dynamics Matter: Both air pollution concentrations and population distribution are not static. Their interaction across space and time is critical for understanding actual exposure risks.
  • Big Data Synergy: The paper demonstrates the power of combining diverse big data sources (mobile phone location, ground sensor networks) to address complex environmental and health challenges.
  • Granular Exposure Assessment: The method allows for near real-time (3-hour) estimation of population exposure at a fine spatial resolution (~1.2 km), enabling more targeted interventions and research.
  • Generalizability: The framework is designed to be adaptable for estimating exposure to other air pollutants (PM10, O3, SO2, NO2), broadening its potential impact.

Key Results

  • The proposed method successfully yields real-time estimations of population exposure to PM2.5 concentrations and cumulative inhaled PM2.5 masses with a 3-hour updating frequency.
  • Comparison with census-based methods revealed significant biases in cumulative inhaled PM2.5 mass estimations (up to over 100%) and population-weighted PM2.5 concentrations (up to ~30 µg/m³) when population dynamics were ignored.
  • The average percentage bias between the MPL-based and census-based estimations was found to be 14.9% (3-h), 5.8% (1-day), 4.7% (1-week), and 3.9% (1-month).
  • The study provides nationwide estimations of population distribution, PM2.5 concentrations, and cumulative inhaled PM2.5 masses at a ~1.2 km resolution.

Limitations and Future Work

  • Assumption of Constant City Population: The method assumes that inter-city mobility does not significantly alter the total population of a city within a short time window. This neglects large-scale human movements and migrations.
  • Data Representativeness: Volunteer-produced geospatial data like MPL records may underrepresent certain population groups (children, elderly, poor) who are less frequent active users. However, the massive volume of data is argued to still provide robust population distribution patterns.
  • PM2.5 Estimation Accuracy: While the GWR model improves PM2.5 estimation, spatial interpolations can still limit accuracy in areas with sparse station data. Future work could focus on improving PM2.5 estimation by integrating satellite data and diurnal change patterns.
  • Simplified Inhaled PM2.5 Model: The model for cumulative inhaled PM2.5 masses simplifies by neglecting outdoor-indoor ratios and variations in inhaled air volume due to age, gender, and activity levels. Future work could refine this by incorporating more detailed individual-level factors.
  • Future Directions: The authors suggest extending the framework to other pollutants, improving PM2.5 estimation accuracy, and further refining the inhaled PM2.5 model. They also highlight the potential for this method to support environmental exposure assessments and policy-driven actions.

Summary updated

Abstract

Extremely high fine particulate matter (PM<sub>2.5</sub>) concentration has been a topic of special concern in recent years because of its important and sensitive relation with health risks. However, many previous PM<sub>2.5</sub> exposure assessments have practical limitations, due to the assumption that population distribution or air pollution levels are spatially stationary and temporally constant and people move within regions of generally the same air quality throughout a day or other time periods. To deal with this challenge, we propose a novel method to achieve the real-time estimation of population exposure to PM<sub>2.5</sub> in China by integrating mobile-phone locating-request (MPL) big data and station-based PM<sub>2.5</sub> observations. Nationwide experiments show that the proposed method can yield the estimation of population exposure to PM<sub>2.5</sub> concentrations and cumulative inhaled PM<sub>2.5</sub> masses with a 3-h updating frequency. Compared with the census-based method, it introduced the dynamics of population distribution into the exposure estimation, thereby providing an improved way to better assess the population exposure to PM<sub>2.5</sub> at different temporal scales. Additionally, the proposed method and dataset can be easily extended to estimate other ambient pollutant exposures such as PM<sub>10</sub>, O₃, SO₂, and NO₂, and may hold potential utilities in supporting the environmental exposure assessment and related policy-driven environmental actions.
BibTeX
@article{2018real,
  author = {Бин Чэн and Yimeng Song and Tingting Jiang and Ziyue Chen and Bo Huang and Bing Xu},
  title = {{Real-Time Estimation of Population Exposure to PM2.5 Using Mobile- and Station-Based Big Data}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2018},
  doi = {10.3390/ijerph15040573},
}
Figure