Predicting Population Adaptation to Climate-Induced Meteorological Variability & Assessing its Potential Effect on Population Exposure to Air Pollution and Health

  • Developing an integrated state of the art population exposure model.

  • Characterising population mobility patterns and activity patterns of population using cellular network mobile phone data from millions of people in New York and Boston.

  • Modelling near surface air temperature in high spatial resolution through a combination of satellite-based observations, meteorological data and spatial predictors. Air quality is modelled through the use of satellite observations, and spatial and temporal covariates.

  • Quantifying how people/populations adapt in terms of their spatio-temporal mobility and time-activity patterns in response to temperature and air pollution changes, with respect to their socio-economic status.

  • Assessing whether climate-related changes in human behavior influence exposure to air pollution and temperature, and subsequent adverse health effects.


Airscapes Singapore – Online Commuter Air Pollution Exposure Assessment Tool Using Data from a Distributed Network of Environmental Sensors LINK TO AIRSCAPES WEBSITE

  • Led the development of an interactive web application to allow commuters to learn how individual experiences lead to varying levels of air pollution exposure metrics.

  • This tool allows individuals to learn how different profiles (gender, age, BMI), commuting routes and activity levels (walking, cycling or jogging) can lead to different exposure levels and inhaled doses of air pollution.

  • Inhaled air pollution doses were based on the following: (1) measured exposure levels specific to a selected route, (2) travelling speed (and time spent exposed), and (3) respiratory parameters which vary for males/females, different ages and for different levels of physical exertion. 

  • Computed exposure metrics based on respiratory parameters appropriate to selected profiles and activity categories. Inhaled CO and NO2 were computed for each of these.

  • Awarded First Prize in the United Nations Environment Program Eye On Earth Summit 2015. 


The Impact of Human Mobility Patterns on Quantifying Population Exposure to PM2.5 and Mortality Effect Estimates in the Greater Boston Area

  • Predicted air pollution (PM2.5) concentration levels using satellite-based aerosol optical depth (AOD) data, a MODIS processing algorithm namely the Multi-Angle Implementation of Atmospheric Correction (MAIAC) and statistical techniques to estimate localized predictions

  • Identified home and work locations of subjects using hundreds of millions of geo-referenced mobile phone usage records from approximately 1.2 million individual phone users for an extensive time period.

  • Computed acute PM2.5 exposures for approximately 400,000 people for different scenarios of exposure.

  • Poisson regression analysis was used to relate exposure estimates for both scenarios to all-cause mortality, cardiovascular- and respiratory-related mortality.


Urban Vehicular Emission Inventory Using Opportunistic Travel Data from Vehicle Fleets and Micro-scopic Emissions Modelling LINK TO ARTICLE IN ATMOSPHERIC ENVIRONMENT

  • Analyzed GPS trajectory data from over 15,000 vehicles with the aim of predicting transportation air pollution emissions for Singapore.

  • Quantified instantaneous velocity and acceleration of vehicles in high spatio-temporal resolution, which provided the basis for the subsequent microscopic emissions model.

  • Estimated carbon dioxide (CO2), nitrogen oxides (NOx), volatile organic compounds (VOCs) and particulate matter (PM) emissions. Identified highly localized areas of elevated emissions levels, with a spatio-temporal precision not possible with previously used methods for estimating emissions.

  • Quantified daily emissions for the total motor vehicle population of Singapore and compared results to another emissions dataset from the National Aeronautics & Space Administration (NASA).

  • Demonstrated that high-resolution spatio-temporal vehicle traces detected using GPS in large transportation fleets could be used to infer highly localized areas of elevated air pollution emissions in cities.


Distributed Network of Mobile Environmental Sensors for Evaluating Personal Exposure to Air Pollution LINK TO AIRSCAPES WEBSITE 

  • Conducted an air quality monitoring campaign using a distributed network of environmental sensors.

  • Recruited and trained ~40 participants to conduct air quality monitoring using portable air quality sensors and smart phones. 

  • Set up network architecture to transfer data collected in sensor devices to a cloud platform, and a local server. Developed web application for displaying the air quality and meteorological data collected in real-time.

  • Determined personal exposure to CO and NO2 for 60 samples collected.

  • Personal exposure as determined by the network of sensors was compared to exposure as determined by the National Environment Agency (NEA) municipal monitors. The exposure surface as determined by the sensors for random locations was also compared to the surface as determined by the NEA district measurement.


‘Exposure Track’ - Quantifying Population-Weighted Exposure to PM2.5 in New York City using Mobile Phone Based Activity Counts LINK TO ARTICLE IN ENVIRONMENTAL SCIENCE & TECHNOLOGY

  • Evaluated population exposure to air pollution considering spatial and temporal variations in air pollution concentration estimates as well as population count estimates in 71 regions of New York City.

  • Deciphered spatio-temporal population distribution patterns using several million counts of connections to the cell-phone network for New York City.

  • Developed air pollution concentration surface plots using New York City Community Air Survey data.

  • A Monte Carlo probabilistic analysis was completed to indicate most likely areas of elevated air pollution levels and population weighted exposures.


Comparison of Predicted Particulate Matter Dose and Heart Rate Variability Response in Cyclists, Pedestrians, Bus and Train Passengers LINK TO ARTICLE IN SCIENCE OF THE TOTAL ENVIRONMENT

  • Investigated the acute relative changes in heart rate variability (HRV) due to personal predicted PM exposures, inhaled doses and lung deposited doses in subject commuters.

  • Recruited 35 subjects for the purposes of this study. The sample consisted of cyclists, pedestrians, bus, train and car passengers.

  • Determined lung deposited doses of air pollution using a numerical model of the human respiratory tract. This numerical model was developed using Matlab software. Previous studies had not accounted for varying ventilation rates between modes, individuals and during commutes.

  • Applied statistical analysis techniques, namely linear mixed models with fixed and random effects, to examine air pollution exposure dose metrics and HRV responses in 122 commutes sampled.

  • Demonstrated that declines in HRV indices (SDNN and rMSSD) due to inter-quartile range increases in PM10 and PM2.5 lung deposited doses were stronger in pedestrians and cyclists, in comparison to bus and train passengers.


Dublin Cycling Study - Evaluating Artificial Neural Networks for Predicting Minute Ventilation and Lung Deposited Dose in Commuting Cyclists LINK TO ARTICLE IN J TRANSPORT & HEALTH

  • Improved air pollution exposure metric estimates for cyclists.

  • Cardiopulmonary parameters (breathing rates and heart rate), personal air pollution exposure, meteorological and GPS data were continuously monitored in 60 cyclist participants while cycling on 7-8km routes in Dublin City Centre.

  • Determined lung deposition both empirically in the field and numerically using a mathematical model.

  • Developed a numerical model of the human respiratory tract using Matlab, and used this to predict air pollution doses in the lungs of commuters.

  • Tested Artificial Neural Networks and other statistical techniques for predicting breathing rates and thus air pollution uptake in the lungs of cyclists using urban environmental (air pollution exposure, meteorological variables) and GPS acquired data (cycling speed, road topography, etc.).