key: cord-0989610-x5ewu49p
authors: Bergmann, M. L.; Andersen, Z. J.; Amini, H.; Ellermann, T.; Hertel, O.; Lim, Y. H.; Loft, S.; Mehta, A.; Westendorp, R. G.; Cole-Hunter, T.
title: Exposure to ultrafine particles while walking or bicycling during COVID-19 closures: a repeated measures study in Copenhagen, Denmark
date: 2021-06-05
journal: Sci Total Environ
DOI: 10.1016/j.scitotenv.2021.148301
sha: fa5da973a6a0639193d9123e2f9025a387f2849c
doc_id: 989610
cord_uid: x5ewu49p

Ultrafine particles (UFP; particulate matter <0.1 μm diameter) emitted from motorized traffic may be highly detrimental to health. Active mobility (walking, bicycling) is increasingly encouraged as a way to reduce traffic congestion and increase physical activity levels. However, it has raised concerns of increased exposure to UFP, due to increased breathing rates in traffic microenvironments, immediately close to their source. The recent Coronavirus Disease 2019 (COVID-19) societal closures reduced commuting needs, allowing a natural experiment to estimate contributions from motorized traffic to UFP exposure while walking or bicycling. From late-March to mid-July 2020, UFP was repeatedly measured while walking or bicycling, capturing local COVID-19 closure (‘Phase 0’) and subsequent phased re-opening (‘Phase 1, 2, 2.1 & 3’). A DiSCmini continuously measured particle number concentration (PNC) in the walker/bicyclist’s breathing zone. PNC while walking or bicycling was compared across phased re-openings and the effect of ambient temperature, wind speed and direction, was determined using regression models. Approximately 40 repeated 20-minute walking and bicycling laps were made over 4 months during societal re-opening phases related to the COVID-19 pandemic, from late March to mid July 2020 in Copenhagen. Highest median PNC exposure of both walking (13,170 pt/cm3, standard deviation (SD): 3560 pt/cm3) and bicycling (21,477 pt/cm3, SD: 8964) was seen during societal closures (Phase 0) and decreased to 5,367 pt/cm3 (SD: 2949) and 8,714 pt/cm3 (SD: 4309) in Phase 3 of re-opening. These reductions in PNC were mainly explained by meteorological conditions, with most of the deviation explained by wind speed (14–22%) and temperature (13–10%). Highest PNC was observed along major roads and intersections. In conclusion, we observed decreases in UFP exposure while walking and bicycling during societal re-opening phases related to the COVID-19 pandemic, due largely to meteorological factors (e.g., wind speed and temperature) and seasonal variations in UFP levels.

Air pollution is a major threat to public health worldwide, and the fourth leading contributor to disease burden worldwide according to the most recent Global Burden of Disease Study (1) . Globally, ambient air pollution is responsible for about 4.5 million premature deaths every year (1) , and 4,200 in Denmark (2) . Extensive research has been conducted especially on the health burden related to particulate matter (PM) of diameter < 2.5 µm (PM 2.5 ) and nitrogen dioxide (NO 2 ) (3, 4) . Due to a lack of regulation and measurements, evidence is lacking on the health effects related to ultrafine particles (UFP; diameter < 0.1 µm).

Toxicological evidence suggests that UFP are potentially more harmful to health than larger particles. This is largely due to their high surface-to-mass ratio and small size, allowing them to penetrate deeper into the lungs, epithelium, and translocate to the blood system and organs while carrying potential toxins via particle surface adsorption (5) .

People experience peaks in daily exposure to air pollution, such as UFP, while they move through city streets, and in transport microenvironments close to motorized traffic. UFP levels in these microenvironments, for different modes of commuting and leisure time J o u r n a l P r e -p r o o f activities including walking or bicycling, may be orders of magnitude higher than those captured by annual average exposures at the home address or daily means of air pollution at city background monitors (6) . Furthermore, the choice of travel modes may significantly affect individual exposure to UFP (7) . With the increasing interest in the promotion of active mobility through walking and bicycling, it is useful to quantify relative exposures to UFP in these travel mode microenvironments (8) . Active mobility increases breathing rates while in close proximity to primary sources of UFP, consequently increasing concentrations of UFP taken into the body (9) . This potential higher pollutant dose can have acute adverse effects such as reductions in lung function due to airway inflammation (10) , discomfort for healthy individuals (11) or worsened symptoms for asthmatic individuals (12) . While physical activity such as that derived through active mobility should be encouraged, air pollution concentrations in this microenvironment should be reduced such as by reducing proximity to or volume of motorized traffic.

The recent Coronavirus Disease 2019 (COVID-19) pandemic provided a natural experiment (or counter-factual) for reduced traffic volumes through the imposition of societal closures in many cities around the world. These closures of workplaces and schools largely reduced the need to commute and therefore decreased motorized traffic-related emissions, including UFP

We opportunistically sampled particle number concentrations (PNC) from the onset of the COVID-19 societal closures implemented in Copenhagen in March to control the COVID-19 pandemic, Denmark. The local closures began mid-March, 2020, with "Phase 0" (March 13 th , 2020). Our sampling began late-March (March 27 th , 2020), in the middle of Phase 0, and continued through to Phase 3 (June, 2020). See Table 1 for an outline of Phases, date ranges, and details of societal closures implemented. The societal closures starting in March coincided with reductions in total, and especially private vehicles", traffic counts in Denmark ( Table 1) . The lowest total counts during societal closures were recorded in week 13 of 2020, 43% lower compared to week 9 in the same year before implementation of the societal closures, after which traffic counts slowly increased again towards reaching pre-closure"s levels in June 2020 (16) . All walking or bicycling trips were performed at approximately the same time of day over three months, between March 31 st and July 16 th , 2020 (from "Phase 0" to "Phase-3" societal re-opening) (17) .

* Changes in traffic counts are given as national, weekly values, compared to week 9, 2020, and are here averaged across phase periods (16) . Only national data was available, which can be seen as indicative of trends in Copenhagen.

Our chosen walking route is a lap around an inner-city lake, which is popular among both walkers and joggers. The south-west part of this route is close to streets with high traffic intensity of around 27,000 vehicles daily (pre-closure) (18) , while the path is further away from traffic in the north-east. The distance of this route is approximately 2 km. Monitoring of the route started within the time period of 12:30-14:30, once per day on weekdays. This time was chosen to cover a typical "after-lunch-walk" around the lakes, which is popular among people working in offices in Copenhagen"s city center.

Our chosen bicycling route is an inner-city lap passing by a regulatory curbside UFP monitoring stationsee Supplemental Figure S1 . This route leads along streets with often For quality control purposes, according to manufacturer recommendation, "zero checks" were performed immediately before and after measurements using a HEPA filter. Further, validation of data was made by week-long co-location at a regulatory air quality monitoring station (H. C. Andersens Boulevard) at the beginning and end of the measurement campaign,

in March and June. The station is equipped with a Scanning Mobility Particle Sizer (SMPS) that counts particles with mobility diameters between 11 and 478 nm every three minutes.

Concurrent geospatial coordinates were recorded with a GPS watch ("Forerunner 920XT";

Garmin Ltd., USA). Measurements were not collected on days or at times with precipitation or high humidity (>90%) according to the manufacturer"s recommendation.

Background concentrations of UFP, coinciding with time of trip performance, were obtained from the regulatory air quality monitoring station (H. C. Andersens Boulevard, Figure S1 ).

Meteorological information, as hourly means, coinciding with trip times, of temperature, relative humidity, wind speed and wind direction, was collected during the study period at a nearby monitoring station (H. C. Ørsted).

First, all time points where the HEPA filter was attached were excluded (33%/32% of total dataset for walking/cycling, respectively) and data with a particle diameter equal to zero (both 7% of total dataset) or PNC above 1 million (both <0.01% of total dataset) were removed for quality control purposes. PNC and diameter were aggregated as means per each trip time and merged with meteorological parameter means at the corresponding time. Next,

Journal Pre-proof PNC while walking or bicycling was described as mean, standard deviation (SD), median, range and interquartile range (IQR) of trips per transport mode.

For validation of UFP data which was measured during bicycling, comparisons were made to a regulatory-grade particle counter within a regulatory air quality monitoring station that is located on the bicycling route (H.C. Andersens Boulevard). Trip means were compared with time-coinciding 20-minute means from the nearby station using Welch Two Sample t-tests.

Further to this, with co-location of the DiSCmini at the same station for several weeks before and following our measurement campaign, we compared hourly means of either instrument using Spearman correlation tests and Bland Altman plots.

Linear models were used to examine the difference in PNC across COVID-19-re-opening phases. In a basic model, adjustment was made for time trend (a numeric variable of the date) and day of week using a Generalized Additive Model (GAM) (21) . In adjusted models, reopening phase, temperature, wind speed and wind direction were included one at each time.

Furthermore, smoothing plots were used to visualize the relationship between PNC and meteorological factors. Degrees of freedom (df) for smoothing terms were determined based on unbiased risk estimation (UBRE) (22) .

Fully adjusted models were constructed as follows:

PNC ~ s(time trend, df=9) + factor(day of week) + factor(phase) + factor(wind direction) + s(wind speed, df=9) + s(ambient temperature, df=9) s is for a smoothing term for time trend, wind speed or ambient temperature.

Medians of all trip data were geospatially aggregated in buffer zones of 20 meters/30 meters for bicycling/walking and visualized using ArcGIS (ESRI).

All statistical (excluding geospatial) analyses were performed in R (version 4.0.2). Statistical significance was accepted at p < 0.05.

Approximately 100,000 seconds of data were collected in total, which were aggregated as means of 41 walking (n=43,600) and 43 bicycling (n=46,200) repeated trips. Figure S2) for Bland Altman plots.

J o u r n a l P r e -p r o o f

Coinciding 20-minute means from the time of bicycling trip performance were compared and seen as not significantly different between the DiSCmini and the regulatory monitor (p = 0.08; t = -1.79; df = 41.2; Figure S3 ).

Mean and median exposure to UFP during trip times was lower for walking compared with bicycling, with mean exposures of 8810 particles per cubic centimeter of air (pt/cm 3 ) while walking and 11,963 pt/cm 3 while bicycling. Mean particle size was equivalent for walking and bicycling at approximately 40 nm ( Table 2 ). Supplemental Table S1 shows summary statistics of the total dataset. Across the study period, our repeated measures showed variation in PNC means across individual days for both walking and bicycling ( Figure 2) . Across COVID-19 societal reopening phases, our repeated measures showed statistically significant differences in PNC mean for bicycling, but not for walking ( Figure 3) . J o u r n a l P r e -p r o o f PNC while walking was highest at the start of study period, during societal closures Phase 0, and decreased monotonically from Phase-0 to Phase-3 (Table 3) . Similarly, PNC while bicycling was highest during Phase 0 and decreased to Phase-2.1, before it increased again in Phase-3. 10.4%). While the raw data shows statistically significant changes in PNC across phases for bicycling, these effects become insignificant after adjustment for covariates ( Figure 3 and Table S2 ). Smoothing plots for the effect of the separate meteorological parameters on PNC showed an inverse relationship between temperature and PNC, whereas wind speed had a less clear, but overall positive effect on PNC (Supplemental Figure S4) .

The levels and variability of meteorological factors, including temperature, wind speed, wind direction, and humidity, across phases are presented in Supplemental Figure S5 . Mean temperature increased from around 11°C in Phase 0 to 21°C in Phase 3, while wind speed decreased from 5 m/s to 4 m/s in the same period. Note different y-axis scales.

In this opportunistic repeated measures study, we found that phased re-opening of society and subsequent change in traffic intensity following COVID-19 closures in Copenhagen in the period of March-July of 2020 did not coincide with increases in UFP exposure while walking or bicycling. The observed gradual reduction in PNC during re-opening phases was to a large extent explained by meteorological parameters, namely wind speed and temperature, indicating that meteorology and seasonal variations were stronger drivers of UFP levels in this study period than COVID-19 societal closure measures and related changes in traffic counts. Moreover, we observed that exposure to UFP while bicycling was higher than while walking, with peak exposure to particle number concentrations being approximately double for bicycling compared to walking. This difference is probably related to differences in the chosen routes for walking and bicycling, as bicycling was performed closer to busy roads with motorized traffic, a main source of UFP. Along both routes, hotspots in UFP exposure were observed due to physical proximity and congestion associated with crossing of intersections (e.g., stopping at traffic signals), which we could visualize using geospatial information.

Several studies have measured UFP exposure while bicycling, mostly in Europe and North

America. Among them, some studies compare exposure at different times of the day (19, (23) (24) (25) (26) (27) (28) (29) or on different routes (high-vs. low-exposure traffic routes) within a city (20, 27, (29) (30) (31) .

Other studies have aimed at comparing exposure for different modes of transportation, such as car, bus, metro, bicycle or walking (19, (32) (33) (34) (35) (36) (37) (38) (39) .

Six studies conducted in Belgium, Canada, Switzerland, China and the UK found higher UFP levels during morning rush-hours compared to afternoon rush-hours (19, (23) (24) (25) (26) (27) . Our observation of lower PNC while walking compared to bicycling could also be time-of-day dependent, with walking measured in the early afternoon (non-rush-hour) compared to bicycling in the afternoon rush-hour. Alternatively, other studies have found increasing levels J o u r n a l P r e -p r o o f of UFP in relation to closer proximity to traffic (20, (29) (30) (31) . This would largely explain the samelower PNC while walking, several meters further from road traffic, compared to higher PNC while bicycling. A recent review comparing UFP exposure in different modes of transportation found that car drivers tend to be most exposed to UFP, followed by people travelling by bus, bicycling and walking (40) .

There is only a single study that has quantified exposure to UFP during commute in rush hour during weekdays in Copenhagen in 2005, on a route comparable to ours, which reported an average exposure of 32,400 pt/cm 3 (41) , which is nearly three times higher than that observed in our study. This difference in levels since 2005 and 2020 may reflect improvements in fuel standards and an increasing share of emission-free and active modes of transportation over the 15 years.

Meteorological factors, mainly wind speed and temperature, have been shown to influence variation in air pollution levels between days and seasons of the year (42, 43) . Both have previously been found to be inversely related to UFP levelsthe increase in temperature and wind speed leads to lower UFP levels (26, 28, 44) . Wind direction can also have an influence on UFP levels (23).

In our study, mean temperature showed an overall inverse relationship with mean PNC, as 

Traffic-sourced PNC emissions were estimated to have reduced by more than 40% following societal closures in China (51) . Other pollutants such as NO 2 , highly correlated with UFP as a traffic-related pollutant, were estimated by satellite monitoring to decrease by up to 20%

following this shutdown in eastern and central China (52) (53) (54) . One study of a neighborhood in the USA found up to 70% less traffic and up to an equivalent reduction in median PNC (13) . Another study investigating traffic-related air pollution, including UFP, while adjusting for meteorology found reductions in UFP significant only when making this adjustment (55) .

They conclude that COVID-19 impacts on traffic-related air pollution levels can differ completely depending on weather scenarios, which means that our findings are not necessarily conflicting with theirs.

Strengths of this study include repeated, real-time measurements using a benchmark personal particle counter sampling immediately within the breathing zone. Another strength is the verification of sampled personal exposure levels through the availability of "background" regulatory monitoring stations, checking: (1) personal particle counter measurement accuracy per cycling trip, and; (2) personal particle counter measurement drift at the beginning and end of the multiple month-long study period per week-long co-location period. Importantly, we were able to make adjustments in our analyses for weather conditions per trip through the availability of corresponding weather station data. (56, 57) . Fourth, this study could not distinguish between different sources of UFP. While traffic is usually the main source of UFP in urban areas, other sources such as non-industrial combustion (e.g., wood stoves) also contribute to local UFP concentrations in Copenhagen, especially in residential areas (58, 59) . This may have influenced the higher levels of UFP in the beginning of the study period, coming out of winter with people using wood stoves more frequently and spending more time than usual at home due to the COVID-19 societal closures, but also the lower levels of UFP going into summer with increasing temperatures.

Finally, our study is novel and limited in representativeness as a commuting exposure study, outside of active mobility for physical activity purposes during work/school closures, to compare to previous commuting exposure studies. The bicycling route is unrealistic as a commute in that it is a loop, however it is comprised of popular paths and indicates exposure levels while performing active mobility. Obtaining physical activity through active mobility J o u r n a l P r e -p r o o f is particularly popular in Copenhagen, where 49% of inhabitants bicycled for their daily commutes to work or education in 2018; increasing from 38% in 2009 (15) . As physical activity is encouraged to promote health benefits including a strengthened immune system, active mobility can be promoted recreationally (e.g., walking or cycling in a loop) to maintain levels no longer achieved through daily commuting.

We observed decreases in UFP exposure during walking and bicycling commutes during societal re-opening phases related to COVID-19 pandemic in the period from late March to mid July 2020 in Copenhagen, Denmark. This decrease was to a large extent explained by meteorological factors (wind speed and temperature), which seemed to be stronger drivers of UFP levels during this period than COVID-19 societal closures.

Future work will investigate peak versus off-peak commuting (with new city UFP monitoring stations), as well as accuracy drift of the DiSCmini device.

Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet

The Danish Air Quality Monitoring Programme

Air quality in Europe -2020 report

The health effects of ultrafine particles

Estimated effects of air pollution and space-time-activity on cardiopulmonary outcomes in healthy adults: A repeated measures study

A review of commuter exposure to ultrafine particles and its health effects

Physical Activity through Sustainable Transport Approaches (PASTA): protocol for a multi-centre, longitudinal study

Wearable Sensors for Personal Monitoring and Estimation of Inhaled Traffic-Related Air Pollution: Evaluation of Methods

Acute respiratory response to traffic-related air pollution during physical activity performance

Utility of an alternative bicycle commute route of lower proximity to motorised traffic in decreasing exposure to ultra-fine particles, respiratory symptoms and airway inflammation -a structured exposure experiment. Environ Health

Respiratory Effects of Exposure to Diesel Traffic in Persons with Asthma

Reductions in traffic-related black carbon and ultrafine particle number concentrations in an urban neighborhood during the COVID-19 pandemic

Population in Denmark

Technical and Environmental Administration, (TMF), Mobility. The Bicycle Account

Trafikkens udvikling i tal

Danish National Police. Political agreements and initiatives | | Coronasmitte.dk

Center for Trafik og Byliv / Centre of Traffic and Urban Life

Commuter exposure to ultrafine particles in different urban locations, transportation modes and routes

Assessment of cyclists' exposure to ultrafine particles along alternative commuting routes in Edinburgh

Generalized Additive Mixed Models using 'mgcv' and 'lme4

Statistical Methods for Environmental Epidemiology with R: A Case Study in Air Pollution and Health

Cyclist exposure to black carbon, ultrafine particles and heavy metals: An experimental study along two commuting routes near

Exposure assessment of a cyclist to PM10 and ultrafine particles

Personal exposure of street canyon intersection users to PM2.5, ultrafine particle counts and carbon monoxide in Central London

The impact of traffic volume, composition, and road geometry on personal air pollution exposures among cyclists in Montreal, Canada

Exposure assessment of cyclists to UFP and PM on urban routes in

Cyclist exposure to UFP and BC on urban routes in

On-bicycle exposure to particulate air pollution: Particle number, black carbon, PM 2.5 , and particle size

Potential pollution exposure reductions from small-distance bicycle lane separations

Inhaled particle counts on bicycle commute routes of low and high proximity to motorised traffic

A travel mode comparison of commuters' exposures to air pollutants in Barcelona

Commuter exposure to fine and ultrafine particulate matter in Vienna

Particulates and noise exposure during bicycle, bus and car commuting: A study in three European cities

Exposure to Particulate Matter Air Pollution Is Affected by Mode of Transport, Fuel Type, and Route. Environ Health Perspect

Variations in exposure to traffic pollution while travelling by different modes in a low density, less congested city

Determinants of Personal Exposure to PM 2.5 , Ultrafine Particle Counts, and CO in a Transport Microenvironment

Exposure to particulate matter in traffic: A comparison of cyclists and car passengers

Exposure to ultrafine and fine particles and noise during cycling and driving in 11 Dutch cities

Comparison of air pollution exposures in active vs. passive travel modes in European cities: A quantitative review

Personal Exposure to Ultrafine Particles and Oxidative DNA Damage. Environ Health Perspect

Fine particulate matter and carbon monoxide exposure concentrations in urban street transport microenvironments

A review of commuter exposure to ultrafine particles and its health effects

Particulate matter exposure along designated bicycle routes in Vancouver, British Columbia

Environmental perspective of COVID-19

COVID-19 pandemic and environmental pollution: A blessing in disguise?

Corona-nedlukningens indflydelse på luftkvaliteten i Danmark i 2020

Changes in air pollution levels after COVID-19 outbreak in Korea

Decrease in Ambient Fine Particulate Matter during COVID-19 Crisis and Corresponding Health Benefits in Seoul

Severe air pollution events not avoided by reduced anthropogenic activities during COVID-19 outbreak

Changes in source contributions to particle number concentrations after the COVID-19 outbreak: Insights from a dispersion normalized PMF

Air pollution reduction and mortality benefit during the COVID-19 outbreak in China

Does lockdown reduce air pollution? Evidence from 44 cities in northern China

COVID-19 as a factor influencing air pollution? Environ Pollut Barking Essex

Impacts of the COVID-19 responses on traffic-related air pollution in a Northwestern US city

Comparability of mobility particle sizers and diffusion chargers

Comparison of the DiSCmini aerosol monitor to a handheld condensation particle counter and a scanning mobility particle sizer for submicrometer sodium chloride and metal aerosols

Kildebidrag til sundhedsskadelig luftforurening i København

Videnskabelig rapport fra DCE -Nationalt Center for Miljø og Energi nr. 57). Available from

Pollution from residential burning: Danish experience in an international perspective

COVID-19 closures reduced the need to commute and therefore local traffic counts

We measured UFP, walking and bicycling, during COVID-19 closures and re-opening

PNC levels decreased continuously during phased re-opening

PNC decreases mainly driven by meteorological factors (wind speed, temperature)

This study was partly funded by the Novo Nordisk Foundation Challenge Programme