1. Questions
Concern about social and spatial inequalities in air pollution exposure and health burdens within populations has prompted researchers to explore exposure during travel and out-of-home activities. Studies are increasingly estimating individual exposure to air pollution using advanced methodologies that can account for the person-specific, dynamic nature of daily activities and travel (Ma et al. 2020; Park and Kwan 2017; Poom, Willberg, and Toivonen 2021). We too have recently generated personal exposure profiles using time-stamped, geo-tagged air pollution data for inhalable (PM10) and fine particulate matter (PM2.5), and Nitrogen Dioxide (NO2) for a sample of 37 participants from Oxford, UK. Data of this nature can support policymakers and local communities involved in environmental governance (Özkaynak et al. 2013).
Additionally, transparent, ethical communication of individual-level exposure data to affected individuals and associated stakeholders has yet to be achieved (Ramírez et al. 2019). This study proposes an online dashboard as an effective medium for such communication, helping to pivot air pollution governance away from reliance on 24-hour or longer-term averages and towards addressing what drives variation and inequality in pollution exposures over space and time. However, the introduction of such a dashboard raises the challenge of geoprivacy: how to respect “individual rights to prevent disclosure of the location of one’s home, workplace, daily activities, or trips”? (Kwan, Casas, and Schmitz 2004, 15). This paper, therefore, addresses the following question: how might an interactive online dashboard represent individual-level air pollution exposure profiles to different audiences?
2. Methods
We collected data on individuals’ daily activity-travel patterns and air pollution exposure using activity-travel diaries, personal air quality monitoring equipment (PAQME) and GPS equipped sensors. 37 individuals based in Oxford, England, participated in the study. Each participant carried a GPS tracker and PAQME for one week and completed an online active-travel diary via a secure web portal at the end of each day. This approach was adopted to confirm the validity of personal daily trajectories. A database compiling individual daily trajectories and time-stamped air pollution measurements was used to estimate individual-level air pollution exposure. Travel routes were connected to street networks using transport mode-specific open route services. The R package osrm (Giraud et al. 2022) was used for private vehicle, walking and cycling routes, and OpenTripPlanner for R (Morgan et al. 2019) for public transport (bus) routes.
An interactive online dashboard to explore individual-level exposure to air pollution has been developed using the flexdashboard package for R Markdown (Sievert et al. 2022). Air pollution concentrations along selected routes are illustrated using the interactive visualisation feature uses leaflet package (Cheng 2022) . The colour of markers corresponds to the pollutant selected, with inhalable particulate matter (PM10 and PM2.5) shown on a scale of cyan to green and Nitrogen Dioxide (NO2) in reddish-brown. The interactive pollution concentrations graph feature uses the plotly (Sievert 2020) package to depict pollutant emissions at specific dates/times.
We have deployed two techniques to respect individuals’ geoprivacy. From available Geomasking techniques for the anonymisation of individual-level data (see Kounadi and Resch (2018) for a review), we use the Geohash technique to aggregate points into areal units (Fox et al. 2013). Geohash length 6 is used to display individual travel-activity trajectories as grid areas (1.2km x 609.4m). Geohash was selected because it is a public geocoding system that encodes location data coordinates into alphanumeric string grid cells using geohashTools for R (Chirico 2020). Furthermore, we designed different access control rules for three categories of user types: researchers, participants, and the public (Table 1). This configuration helps researchers to share appropriate information with different user groups.
3. Findings
The dashboard interface (https://wondolee.github.io/JFF) presents patterns in daily exposure to PM10, PM2.5 and NO2 at different resolutions for different user groups (Table 1). All user groups see three linked displays: on the left a map of Oxford, and on the right a graph with (average) measured pollutant concentrations over the diurnal cycle plus a bar to adjust the time window for which exposure levels are displayed. Study participants are offered full access to their own data (Figures 1a, 2a).[1] In contrast, to protect participants’ geoprivacy, maps and graphs for the public depict how individual-level exposure data aggregated into grid cells and averaged to each hour of the day and across the seven days (Figures 1b, 2b).
To facilitate comparison and examination of health risks that air pollution exposure poses, the graphs in Figure 2 show additional information. First, a trend line indicating pollution concentrations measured at the official UK Automatic Urban and Rural Network (AURN) monitoring station in Oxford’s city centre at St Ebbes street (ID: UKA00518) is depicted. Secondly, the WHO guideline for the 24-h mean NO2 value of 25μg is included (see also Table 2). The differences between PAQME generated air pollution measurements and AURN monitoring station data reflect differences in spatial position, and potentially the larger margin for error associated with PAQME measurement. PAQME generated measurements are best considered indicative of experienced exposure because the performance of low-cost portable sensors can vary under different urban conditions (Lewis and Edwards 2016; Ma et al. 2020). We nonetheless believe that using of such sensors is acceptable when the aim is to convey information on patterns of exposure and induce reflexivity about when and where action to reduce exposure might be considered.
In short, the developed online dashboard can disseminate and exchange variations in air pollution exposure safely by tailoring permission levels to different audiences with advanced de-identification techniques. It paves the way for customisable advice about individual-level air pollution exposure reduction by allowing individuals, citizen collectives, NGOs and policymakers to engage with (their own) daily activity-travel patterns in meaningful ways that preserve Geoprivacy. For instance, it can provide targeted insight for to help active transportation users seeking to reduce total exposure during their commuting or when travelling along main roads during peak hours, if dashboard is regularly updated.
ACKNOWLEDGEMENTS
This research is part of the ‘Personal air pollution exposure assessment using smart sensing technologies’ project (reference: 0008971), funded by the University of Oxford’s John Fell Fund.
Map (Figure 1a) and graph (Figure 2a) for study participants were generated by using author’s data to protect Geoprivacy of participants’.