Next Article in Journal
Traditional Masculinity and Aggression in Adolescence: Its Relationship with Emotional Processes
Next Article in Special Issue
Evaluating the Performance of Low-Cost Air Quality Monitors in Dallas, Texas
Previous Article in Journal
Taking Up and Terminating Leisure-Time Physical Activity over the Life Course: The Role of Life Events in the Familial and Occupational Life Domains
Previous Article in Special Issue
Indoor Air Quality in Domestic Environments during Periods Close to Italian COVID-19 Lockdown
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Using Low-Cost Sensors to Assess Fine Particulate Matter Infiltration (PM2.5) during a Wildfire Smoke Episode at a Large Inpatient Healthcare Facility

1
Environmental Health Services, BC Center for Disease Control, Vancouver, BC V5Z 4R4, Canada
2
Department of Mechanical Engineering, The University of British Columbia, Vancouver, BC V6T 1Z4, Canada
3
Air Health Sciences Division, Health Canada, Ottawa, ON K1A 0K9, Canada
4
Provincial Health Services Authority, Vancouver, BC V6H 4C1, Canada
5
Vancouver Coastal Health, Vancouver, BC V5Z 1A1, Canada
6
School of Population and Public Health, The University of British Columbia, Vancouver, BC V6T 1Z3, Canada
*
Author to whom correspondence should be addressed.
Int. J. Environ. Res. Public Health 2021, 18(18), 9811; https://0-doi-org.brum.beds.ac.uk/10.3390/ijerph18189811
Submission received: 5 August 2021 / Revised: 13 September 2021 / Accepted: 16 September 2021 / Published: 17 September 2021
(This article belongs to the Special Issue Low-Cost Sensors for Environmental Research and Public Health)

Abstract

:
Wildfire smoke exposure is associated with a range of acute health outcomes, which can be more severe in individuals with underlying health conditions. Currently, there is limited information on the susceptibility of healthcare facilities to smoke infiltration. As part of a larger study to address this gap, a rehabilitation facility in Vancouver, Canada was outfitted with one outdoor and seven indoor low-cost fine particulate matter (PM2.5) sensors in Air Quality Eggs (EGG) during the summer of 2020. Raw measurements were calibrated using temperature, relative humidity, and dew point derived from the EGG data. The infiltration coefficient was quantified using a distributed lag model. Indoor concentrations during the smoke episode were elevated throughout the building, though non-uniformly. After censoring indoor-only peaks, the average infiltration coefficient (range) during typical days was 0.32 (0.22–0.39), compared with 0.37 (0.31–0.47) during the smoke episode, a 19% increase on average. Indoor PM2.5 concentrations quickly reflected outdoor conditions during and after the smoke episode. It is unclear whether these results will be generalizable to other years due to COVID-related changes to building operations, but some of the safety protocols may offer valuable lessons for future wildfire seasons. For example, points of building entry and exit were reduced from eight to two during the pandemic, which likely helped to protect the building from wildfire smoke infiltration. Overall, these results demonstrate the utility of indoor low-cost sensors in understanding the impacts of extreme smoke events on facilities where highly susceptible individuals are present. Furthermore, they highlight the need to employ interventions that enhance indoor air quality in such facilities during smoke events.

Graphical Abstract

1. Introduction

Wildfire smoke may become the dominant source of exposure to fine particulate matter (PM2.5) in western North America, which has uncertain human health implications [1]. Short-term exposure to wildfire smoke is associated with an increased risk of acute respiratory outcomes, such as exacerbations of asthma and chronic obstructive pulmonary disease (COPD) [2,3,4,5,6]. Within Canada, short-term exposures during the 2013–2015 and 2017–2018 wildfire seasons were associated with 54–250 deaths due to acute cardiorespiratory outcomes, while long-term exposures were associated with 570–2500 deaths due to chronic disease [7]. Although such estimates are typically made using PM2.5 ambient air quality data, these outdoor concentrations do not explicitly assess variability in indoor PM2.5 exposure. Most people in North America spend at least 85% of their time inside a home or other building [8], meaning that most health effects associated with ambient smoke are due to exposures that occur indoors. As wildfire seasons are expected become longer and more intense [9], it is critical that we study smoke infiltration and effective interventions, particularly in settings that house susceptible populations [10].
The indoor infiltration of outdoor wildfire smoke has not been comprehensively studied, so it remains unclear how much protection remaining indoors provides during wildfire smoke episodes. In addition, there are limited data for multi-story non-residential buildings such as offices and healthcare facilities. Healthcare facilities are particularly concerning because they serve individuals who are more susceptible to smoke exposure due to compromised health status [3,11]. One study during the 2018 Camp Fire may provide some insight. Pantelic et al. (2019) compared two large commercial buildings, one with a mechanical ventilation system (HVAC) that was outfitted with two-staged particle filtration, and one without. The building with HVAC had a mean indoor PM2.5 concentration of 21 µg/m3 and an indoor-to-outdoor ratio of 0.27. In comparison, the building with natural ventilation had a mean indoor concentration of 36 µg/m3 and an indoor-to-outdoor ratio of 0.67 [12]. The latter is more consistent with infiltration coefficients typically reported for single-family residences, which rarely have outdoor air intakes to filter incoming air, and rely on natural ventilation and leaks in the building envelope for air exchange. For example, the mean infiltration factor for 17 homes in southern British Columbia, Canada during the 2004–2005 wildfire seasons was 0.61 [13].
Differences between buildings suggest that airtightness, filtration and ventilation play important roles in outdoor PM2.5 infiltration during wildfire smoke episodes [12]. A more airtight envelope reduces smoke penetration through cracks and crevices, and mechanical ventilation typically provides at least some filtration of incoming outdoor air. Furthermore, it may be possible to increase filtration of fresh or recirculating air by upgrading and/or reducing air bypass around filters. It is also possible to reduce the fresh air intake on some HVAC systems during smoke episodes, though this is antithetical to best practice for COVID-19 risk reduction [14]. Even when measures are taken to reduce smoke infiltration, indoor PM2.5 concentrations vary widely between different buildings and residential homes during smoky periods [15,16]. Methods that evaluate infiltration on a building-by-building basis may be needed to inform actions taken by facility operators and HVAC specialists, especially for facilities where susceptible individuals reside. Low-cost sensors may have an important role to play in evaluating and optimizing building-specific interventions in both existing buildings and new construction projects. Notably, the building stock for hospitals, rehabilitation centers, long-term care, residential care, and publicly-funded housing is highly variable, with many older facilities constructed under outdated building codes. There is currently no comprehensive information about wildfire smoke infiltration in such environments.
Another factor underlying infiltration is meteorological conditions, such as precipitation, wind speed, wind direction, temperature, and relative humidity. Some studies have reported that wind speed, temperature, and ambient relative humidity are negatively correlated with PM2.5 concentrations and the PM2.5 indoor–outdoor ratio [17,18,19,20]. Precipitation is also inversely proportional to PM2.5 concentrations, due to the removal effect where particulate matter is removed from the air to the surface by rain [17]. However, there are some studies that show no correlation or negligible effects of temperature and precipitation [18,21,22], while others show positive relationship between temperature and PM exposure, due to increased window opening for ventilation and cooling [23].
There are many factors affecting infiltration of wildfire smoke, and there is a real opportunity to use low-cost sensors to better understand the impacts on specific buildings. One caveat is that the data quality from low-costs sensors can be uncertain because they are sensitive to many variables, including relative humidity, temperature, particle morphology, and particle composition [24]. Correcting for these effects is necessary to yield higher-quality data for analysis. When evaluated under outdoor field conditions, low-cost PM2.5 sensors are generally well-correlated with reference instruments, but they tend to over-predict concentrations in both ambient and wildfire smoke conditions [25]. Nonetheless, linear correction has been effective for reducing errors [25,26]. Although their performance indoors has not been widely evaluated, their outdoor performance suggest that indoor networks of low-cost sensors can significantly improve our understanding of indoor air quality and its health impacts, especially during significant outdoor air quality events [27].
British Columbia experienced extreme wildfire seasons in 2017 and 2018, including episodes of prolonged smoke that affected multiple healthcare facilities. In response, local health authorities began planning to outfit three inpatient healthcare facilities with multiple indoor and outdoor low-cost PM2.5 sensors. This initiative was intended to support and inform future planning, design, and operational measures required to reduce the health risks associated with outdoor air quality for patients and staff in existing and future facilities. Sensors were installed at a rehabilitation facility in Vancouver in August 2020, with the intention of leaving them in place for at least one year. About 2 weeks later, dense smoke was transported into southern British Columbia from wildfires in California, Oregon, and Washington states. As such, we have a unique opportunity to examine the impacts of wildfire smoke on an inpatient rehabilitation facility.
In this study, we describe methods used to calibrate the low-cost sensor data using measurements from the surrounding regulatory network. After data calibration, we summarize indoor and outdoor PM2.5 concentrations at the healthcare facility during the smoke episode and during typical days on either side of the smoke episode. Specifically, we remove indoor peaks from the indoor measurements, calculate the coefficient of infiltration at each indoor location, and examine the speed of infiltration into the facility.

2. Materials and Methods

2.1. Study Context and Period

Greater Vancouver is a major urban center on the southwestern coast of British Columba, Canada. In 2020, its population was approximately 2.5 million people. The region typically has excellent ambient air quality, with an annual average PM2.5 concentration of approximately 5.0 µg/m3. Days exceeding the 24-h objective of 25 μg/m3 are rare, except during wildfire seasons when the region can experience markedly higher concentrations of PM2.5 originating from fires in British Columbia and other parts of western North America. Our analyses cover the period from August 21 through 31 October 2020. For all analyses we defined the smoke episode as September 8 through 18, which corresponds with the timing of the 11-day air quality advisory issued by Metro Vancouver. We defined the typical days as August 21 through September 7 and September 19 through October 31.

2.2. The Rehabilitation Facility

The facility is a large multi-story rehabilitation center that usually serves inpatient and outpatient populations, but all outpatient services were suspended during the study period due to the COVID-19 pandemic. The facility provides services in four different programs: (1) acquired brain injury; (2) spinal cord injury; (3) arthritis; and (4) neuromusculoskeletal. The main building was constructed in 1972. It has four floors above ground and a basement, and most rehabilitation activities occur on the ground floor in the physical and occupational therapy rooms. However, most therapy programs were offered in patient rooms during the study period in accordance with the pandemic safety plan. Offices and patient rooms are located on the second to fourth floors, and the basement serves a range of operational functions, including the loading bay.
Heating for the facility is provided by radiators, and cooling is provided by an HVAC system with ten air handling units. Most of the windows are fully operable. During the smoke episode the pre-filters in the air handling units were changed more frequently, the cooling system was operational, and building occupants were asked to keep the windows closed. No other special precautions were taken to limit indoor smoke impacts. Due to COVID-19 protocols, entrance to the building was only possible through two doors: one in the main lobby for patients and visitors, and one at the side of the ground floor for staff.

2.3. Low-Cost Sensors

The 2018 model of the Air Quality Egg (EGG) by Wicked Device was used for this study (version 2.0) [28]. All the EGGs had dual Plantower PMS5003 (2016) sensors that measure PM2.5 concentrations via light scattering, with the reported value being the average of the two sensors. The Plantower sensors report an effective range between 0–500 µg/m3 and a maximum consistency error of ±10% at 100–500 µg/m3 and ±10 µg/m3 at 0–100 µg/m3. Each EGG also had a temperature and relative humidity sensor. Some EGGs used in the study also had sensors for carbon dioxide (CO2), nitrogen dioxide (NO2), and volatile organic compounds (VOC), but these data are not included in the analyses. The EGGs log measurements from all sensors at 1-min intervals and automatically upload the data to a central repository.

2.4. Sensor Deployment at the Facility

Nine EGGs were deployed to the facility in August 2020, with two placed outdoors and seven placed indoors at a height of approximately two meters. One outdoor EGG was placed on the roof of the main building, and one was placed outside of a window on the second floor. The latter EGG failed early in the study period and was excluded from the analyses. Three indoor EGGs were placed near the elevator doors in the basement, in the ground floor lobby, and on the fourth floor. The remaining EGGs were placed in a basement office, the ground floor therapy room, a fourth-floor office, and a fourth-floor patient room (Figure 1). The placement of the nine sensors was discussed with subject matter experts to identify locations that would provide useful information for the study objectives.

2.5. Collocation of the Low-Cost Sensors with Federal Equivalency Methods (FEM)

Six EGGs from the larger project were co-located at the Kensington Park ambient air quality monitoring station (Figure 2) from 17 July through 30 July 2020. This station is part of the National Air Pollution Surveillance (NAPS) network that measures PM2.5 using beta attenuation monitors that meet Federal Equivalency Method (FEM) standards. Two of the co-located EGGs (EGG 4 and EGG 6) were later deployed to the facility, in the ground floor lobby and outside on the second floor, respectively. Again, EGG 6 failed early in the study period and was excluded from the facility analyses.

2.6. Data Calibration with a Regional Baseline Model

The average PM2.5 concentrations at Kensington Park in July 2020 were low and did not represent concentrations during the smoke episode, so we developed and applied a novel approach using data from FEM monitors located at the Vancouver International Airport (YVR) and Clark Drive NAPS stations during the study period. These stations are located to the southwest and northeast of the facility, respectively (Figure 2). To check the accuracy of the calibration model it was applied to the six EGGs that were co-located with the FEM monitor at Kensington Park.
All data management, analyses, and visualization for this study were conducted using R version 4.0.3 [29]. Measurements from the EGGs and NAPS stations were available at 1-min resolution, which were smoothed to 15-min intervals using the timeAverage function of the openair package [30]. The following measurements were collected from each EGG for this study: Temperature (T, °C); relative humidity (RH, %); and PM2.5 (µg/m3).
The EGG PM2.5 data collected at the facility rooftop were used to train a calibration model with the average baseline of the FEM PM2.5 concentrations (i.e., estimated regional PM2.5 concentration) in a multiple linear regression that also included temperature, relative humidity, and dewpoint parameters, as suggested elsewhere [31]. The EGG dew point was computed by the Magnus formula [32] as shown in Equation (1):
D P ( T ,   R H ) = λ   · ( ln ( R H 100 ) + β   · T λ + T ) β ( ln ( R H 100 ) + β   · T λ + T )  
where λ = 243.12 °C; β = 17.62; T = temperature measured by the EGG; and RH = relative humidity measured by the EGG.
The data for the calibration model were taken from a 5-week period, from August 21 at 08:00 until September 24 at 08:00. This period was chosen because the rooftop EGG data were relatively continuous throughout. A Kalman smoothing function was applied to the rooftop EGG PM2.5 and FEM PM2.5 data to impute any missing values. The baseline for each time series was then computed using a Rolling Ball algorithm [33], and the average of the FEM baselines was taken. The baselines were computed using the baseline package in R [33]. Assuming the average of the FEM PM2.5 concentrations to be the true value, EGG measurements of PM2.5, temperature (T), relative humidity (RH) and dewpoint (DP) were used to predict the FEM baseline and the resulting equation (Equation (2)) was applied to all EGG PM2.5 data collected at the facility.
Corrected PM2.5 = −40.562 + 1.019*PM2.5 − 1.7102*T − 0.534*RH + 2.153*DP
To test the accuracy of the model, PM2.5 baselines for the FEM collocation period were computed in the same manner as described for the facility datasets. The regional baseline model was then applied to the collocation EGG data and compared with the FEM PM2.5 concentrations. Metrics to quantitatively assess the performance of the calibration model included Pearson r, the mean absolute error (MAE) and the coefficient of variation of the mean absolute error (CvMAE), as calculated by Equation (3). These metrics are used to quantify the correlation between the estimated and true concentration of PM2.5 (Pearson r), the average of differences between the estimated and true concentration of PM2.5 (MAE), and the mean absolute error normalized to PM2.5 concentration (CvMAE). A higher Pearson r signifies a strong linear relationship, and lower MAE and CvMAE signify closer agreement between the calibrated and true values of PM2.5.
C v M A E = 1 a v g . F E M [ P M 2.5 ] × ( 1 n i = 1 n | c a l i b r a t e d [ P M 2.5 ] i F E M [ P M 2.5 ] i   | )
where n = the number of observations in the collocation dataset.

2.7. Removal of Indoor-Generated PM2.5

Before calculating the infiltration coefficients, we used an algorithm to remove indoor peaks that could have resulted from indoor sources, such as kitchen and manufacturing activities in the facility. The algorithm we used was adapted from a previous paper on infiltration of outdoor PM2.5 using 1-h averages [34]. To adapt the approach for 1-min measurements, we modified the algorithm and implemented stricter thresholds to identify indoor peaks. Each indoor peak comprises rising concentrations and decaying concentrations, which were identified as described below.
First, for every time point (t), we checked whether (1) the indoor–outdoor ratio was greater than 2.0, (2) the indoor concentration was at least 2 µg/m3, and (3) the indoor concentration had increased from the previous time point (Equation (4)). Outdoor concentrations were taken from the rooftop sensor at the facility. When these conditions were met, the beginning of an indoor peak was identified, and the data point was removed. Second, we checked all measurements following the beginning of an indoor peak and removed them if the concentration continued to rise (Equation (5)). Finally, we removed the decaying side by checking whether (1) the previous measurement had been removed, (2) the indoor–outdoor ratio was greater than 1.0, and (3) the indoor concentration had decreased from the previous time point (Equation (6)).
I t O t > 2   and   I t 2 μ g / m 3   and   I t I t 1 > 1
I t 1   was   previously   removed   and   I t I t 1 > 1
I t 1   was   previously   removed   and   I t O t > 1   and   I t I t 1     1
where: I = the indoor concentrations at time intervals t and t − 1; O = the outdoor concentration at time interval t.

2.8. Smoke Infiltration

The infiltration coefficient represents the proportion of outdoor PM2.5 that has infiltrated and persisted indoors. Once the indoor peaks were removed, the infiltration coefficient was quantified using distributed lag linear models and the dlnm package [35,36]. In this model, the indoor PM2.5 concentration at time (t) is estimated by the cumulative lagged effects of the outdoor rooftop PM2.5 concentration over a 60-min interval (Equation (7)).
I t = β 1 O t 1 + β 2 O t 2 + + β m O t m +   α   m = 1 ,   2 ,     , 60
where: I t = the indoor PM2.5 concentration at time t; O t m = the outdoor concentration at time t-m minutes; β m = the lagged coefficient of O t m ; α = the intercept.
Each coefficient represents the contribution and persistence of the outdoor concentration at every previous minute over a 60-min interval on the current indoor PM2.5 concentration. The infiltration coefficient (Finf) is quantified by the sum of the lagged coefficients (Equation (8)). We used the 1-min calibrated indoor EGG PM2.5 with peaks removed and rooftop EGG PM2.5 for the calculations during periods that capture the smoke episode and typical days. Missing values were omitted.
F i n f = β 1 + β 2 + + β m   m = 1 , 2 , , 60
To qualitatively examine the persistence of indoor smoke, we used stacked heat maps of EGG PM2.5 concentrations on the facility rooftop and indoors. Each plot shows the smoke period compared with the pre- and post-smoke periods, by hours of the day.

2.9. Meteorological Conditions

The EGGs measure temperature and relative humidity, so we were able to examine differences in these parameters at the facility during the smoke episode and the typical days. Although we could not measure complete meteorology at the facility, we were able to access information on temperature, humidity, wind speed, wind direction, and precipitation at the FEM sites at YVR, Clark Drive, and Kensington Park. We compared meteorological parameters during the smoke episode and the typical days.

3. Results and Discussion

3.1. Calibration Model Performance

Prior to any analyses with the EGG data collected at the facility, they were calibrated using the regional baseline model described above (Equation (2)). The magnitudes and directions of the model coefficients were comparable with those previously reported for similar models [31]. When tested on the collocation period data, Pearson r values ranged from 0.73–0.83 among the EGGs, MAE ranged from 2.61–4.88 µg/m3, and CvMAE ranged from 0.48–0.91 (Table 1). The aggregate performance of the calibration models was relatively strong, with mean Pearson r, CvMAE, and MAE values of 0.80, 0.58, and 3.10, respectively. The average concentration at the Kensington Park ambient air quality monitoring station during the collocation period was very low (4.47 µg/m3), which accounts for the high CvMAE values during our calibration model assessment. The EGGs also reported very low average PM2.5 levels during this period (Table 1).
The 24-h average of uncalibrated PM2.5 concentrations during the typical days and smoke episode were 8.7 µg/m3 and 73.1 µg/m3, respectively In comparison, the 24-h average of calibrated PM2.5 concentrations were 7.5 µg/m3 and 72.0 µg/m3, respectively. The preliminary results of our regional baseline regression have implications for future low-cost sensor calibration studies. When collocation data are not available or representative of the outdoor conditions at the time the study is performed, we suggest that a calibration model can be developed from a regional PM2.5 baseline that reflects the surrounding region during the study period. Use of regional baseline calibration may drastically reduce the need for lengthy collocation periods when performing measurements of pollutants from outdoor sources with low-cost sensors, especially for citizen science projects that require large and distributed sensor networks. We recommend this as an area for future study, specifically as it applies to extreme air quality events such as wildfire smoke episodes.

3.2. Indoor Peak Removal

We applied an algorithm that removed peaks caused by indoor PM2.5 sources at the facility, and the number of peaks varied by location (Table 2). The ground floor therapy room and a patient room on the fourth floor had the most indoor peaks. The therapy room had known indoor sources such as a model kitchen, a facility for manufacturing assistance aids, and supplies for hot wax therapy, all of which were used by staff during working hours. Indoor sources of PM2.5 in the patient room are unknown.

3.3. Indoor and Outdoor PM2.5 Concentrations

Air quality in greater Vancouver was excellent on the typical days during the study period. The 24-h FEM PM2.5 averages (range) at Kensington Park, Clark Drive, and YVR were 4.8 (1.0–17.6), 7.1 (1.5–46.2), and 4.5 (0.9–14.3) µg/m3, respectively (Figure 3). Higher concentrations at Clark Drive reflect its roadside location on a heavily trafficked trucking route. In comparison, the calibrated 24-h PM2.5 average at the facility rooftop was 7.5 (0.0–46.8) µg/m3 (Figure 4). During the wildfire smoke episode, the rooftop EGG at the facility had a 24-h average of 72.0 (7.7–141.6) µg/m3. In comparison, the FEM monitors at Kensington Park, Clark Drive, and YVR had 24-h averages of 76.1 (7.7–172.4), 75.8 (11.1–161.4), and 69.1 (9.5–144.9) µg/m3, respectively, during the smoke episode.
On average, the indoor PM2.5 concentrations at the facility were substantially higher during the wildfire smoke episode than on the typical days, with an average 24-h value of 29.6 µg/m3 compared with 2.4 µg/m3, respectively. The indoor PM2.5 concentrations varied by location, with higher concentrations observed near entrances and exits (Table 3 and Figure 5). During both the smoke episode and the typical days, indoor PM2.5 concentrations were, on average, lower than the outdoor concentration on the rooftop, which is consistent with other studies conducted in large buildings [12,37].

3.4. Indoor Infiltration of Outdoor PM2.5

Infiltration of outdoor PM2.5 was generally higher during the wildfire smoke episode than on typical days, but it varied by location (Table 2). On typical days, the mean (range) infiltration was 0.32 (0.22–0.39) across all indoor locations, and infiltration was highest at the basement elevator and lowest in the ground floor therapy room (Table 2). During the smoke episode, the mean infiltration increased to 0.37 (0.31–0.47), an average increase of 19%, ranging from a decrease of 3% in the fourth-floor patient to an increase of 41% ground floor therapy room. Although the average infiltration was higher during the smoke episode than on typical days, the patient room on the fourth floor observed no substantial change in infiltration during the smoke episode, after censoring indoor-generated peaks (Table 2). The higher infiltration in the patient room during typical days could be due to occupants opening the windows.
Overall, the coefficients suggest that there is more infiltration of PM2.5 when the outdoor concentrations are very high. However, the mean infiltration into the facility during the wildfire smoke episode was considerably lower than reported mean wildfire smoke infiltration for residential homes, which has ranged from 0.56 to 0.79 in previous studies [13,34,38]. This could suggest that the facility building is more protected against infiltration than private residences, or that management of the building during the COVID-19 pandemic helped to protect it. Even so, indoor concentrations at the facility were high compared with benchmarks such as the air quality objective for Metro Vancouver, which is a 24-h average PM2.5 concentration less than 25 µg/m3. If the provincial Air Quality Health Index (AHQI) is applied, any 1 h concentration greater than 60 µg/m3 is in the high risk category [39], including: 50 h near the basement elevator during the 264 h smoke episode; 48 h in the lobby; 14 h in the fourth floor patient room; 11 h in the basement office; and 3 h in the ground floor therapy room.
The infiltration coefficients calculated for the facility are more consistent with results from previous studies conducted in similar large buildings during an extreme wildfire episode. For example, Pantelic et al. reported a value of 0.27 for a large commercial building with HVAC in Berkeley, California when the median hourly outdoor concentration was 21 µg/m3. Similarly, Wheeler et al. reported a value of 0.31 for a public library in Port Macquarie, Australia when the mean outdoor concentration was 30.7 µg/m3 [12,37]. Our infiltration coefficients are also consistent with another study that calculated a median infiltration of 0.45 among commercial buildings in Oregon and California during the September 2020 smoke episode [40]. However, the mean outdoor PM2.5 concentration was not reported for this study. Guidelines from the Canadian Standards Association (CSA) suggests that occupational therapy and physiotherapy health facilities should maintain 6–9 total air changes per hour (ACH) [41], which might make them more susceptible to smoke infiltration if the incoming air is not effectively filtered. The ACH for the building with HVAC in the Berkley study was below 0.30 [12], making it very airtight in comparison. We were not able to measure ACH for this study.
Infiltration of PM2.5 was non-uniform across locations in the facility during both the wildfire smoke episode and typical days. Certain locations, such as the basement elevator and the lobby, were more susceptible to infiltration, resulting in higher coefficients during both periods (Table 2). These differences between locations were not associated with the floor on which the EGGs were located, as there were higher and lower values in the basement, on the main floor, and on the fourth floor. Very few studies have reported on PM2.5 infiltration into multi-story buildings, so it is difficult to evaluate whether these findings are typical. Another study with indoor and outdoor pairs of light-scattering sensors placed at different heights on an 8-story building found differences in concentrations by height, but strong correlation between indoor and outdoor values regardless of height [42]. We would have been able to do more similar analyses if the outdoor EGG on the second story had not failed.
We believe higher infiltration values near the basement elevator and in the lobby are best explained by activities occurring in these two locations. The elevators in the basement are near to the loading bay doors and another service door frequently used by cleaners and staff to access the outside of the building. The EGG by the elevators in the lobby was near the main entrance, which includes two sets of double sliding doors and a 12-foot vestibule between. Both sets of doors are often open at the same time to accommodate patients using mobility aids. Other locations, such as the ground floor therapy room, were better protected against infiltration. While the ground floor therapy room is affected by indoor generation of PM2.5, it also has a dedicated air handling unit that may help to dilute the air more quickly than in other areas of the building. When indoor locations were ranked by infiltration coefficient during the smoke episode and typical days, the orders were similar (Table 2), possibly because they always have relatively similar rates of airflow. These results suggest that indoor infiltration of outdoor particles varies consistently by location within the building, and that infiltration on typical days can help to assess potential infiltration during smoke episodes, to help prioritize occupied locations for additional measures such as deploying sufficiently sized portable air cleaners. This finding also has implications for space design and functional programming developed for new construction of healthcare facilities.
Infiltration of PM2.5 during the smoke episode at the facility may have been mitigated by smoke-specific protocols and restrictions implemented for the COVID-19 pandemic. The pre-filters in the air handling units were changed more frequently during the smoke episode, the cooling system was operational, and building occupants were asked to keep the windows closed. In addition, the COVID-19 safety plan limited building entrances to two doors instead of the usual eight. All patients used the sliding double doors in the ground floor lobby and staff used a different, smaller door at the side of the building. In general, local-level interventions such as portable air cleaners for areas that have higher infiltration might improve conditions without the need for a larger-scale intervention. For example, enforced use of the double door vestibule during smoky conditions could further limit infiltration into the main lobby. Likewise, those without mobility aids could be directed through smaller doors.

3.5. Meteorology during Smoke Episode and Typical Days

There was little to no precipitation during both the smoke episode and the typical days (Table 4). Outdoors, the wind speed and relative humidity were, on average, lower during the smoke episode than during the typical days, though the temperatures were somewhat higher. The average wind direction was consistent across both periods and typical for the coastal region. Indoors, both temperature and relative humidity were higher during the smoke episode than during the typical days. These small differences in outdoor and indoor conditions may have had small effects on the infiltration coefficients based on prior literature [17,18,19,20,23].

3.6. Indoor PM2.5 Patterns

The daily heat map (Figure 5) shows the overall pattern of the smoke episode. We found that the lag between rooftop PM2.5 concentration changes and subsequent indoor changes was short, occurring within 1–2 h. When outdoor PM2.5 concentrations increased sharply at the facility rooftop on the morning of September 11, indoor PM2.5 concentrations across all locations increased within the same time frame, though the impact varied by location. This is consistent with recent reports of rapid outdoor PM2.5 infiltration, where approximately half of the total penetration occurred within the first hour [43]. When rooftop PM2.5 concentrations decreased, indoor levels also changed quickly to reflect outdoor conditions. Indoor PM2.5 concentrations began to decrease quickly after the rooftop PM2.5 concentrations began to drop at approximately 18:00 on September 18. This suggests that outdoor PM2.5 changes, whether increases or decreases, affected indoor air quality almost immediately, but not uniformly (Figure 5). This highlights the utility of air quality advisories and wildfire smoke forecasting as triggers to implement building-specific smoke readiness plans, as recently recommended by ASHRAE [44].

3.7. Limitations

This study has several limitations. First, it was conducted during the COVID-19 pandemic and its associated building restrictions, such as reduced points of entry, suspension of outpatient programs, and no inpatient visitations. These restrictions may have lowered infiltration, and we do not know how smoke would have infiltrated during an extreme episode under normal circumstances. Even so, some of these restrictions may provide insight into protecting large buildings from wildfire smoke infiltration through simple measures such as limiting the number of entrances. Second, the indoor EGGs were not co-located with indoor FEM monitors, and we had to calibrate them with outdoor data, which are not representative of conditions inside the facility. Third, we may not have removed all indoor peaks, and there may be some remaining indoor contribution that is not captured by the algorithm that identified peaks. The relative contribution of indoor sources may have been higher during the typical periods, potentially reducing estimates of infiltration compared with the wildfire smoke episode. During the wildfire smoke episode, both indoor and outdoor concentrations were much higher, and any indoor background concentrations would have made a smaller relative contribution to total indoor PM2.5, so the infiltration coefficient may have been attenuated.

4. Conclusions

This study is the first to evaluate wildfire smoke infiltration into a healthcare facility, where people with compromised health status reside. We found that infiltration during an extreme wildfire smoke episode at the facility was, on average, 19% higher than infiltration during typical days. Indoor concentrations increased across all locations during a smoke episode, suggesting that no locations were completely protected from smoke, though infiltration was higher in areas near to the limited entrances and exits in use during the COVID-19 pandemic. Restricting entrances and using double doors with vestibules may help minimize overall smoke infiltration. Additionally, the indoor air at the facility quickly reflected outdoor changes measured by the rooftop EGG, whether PM2.5 concentrations were increasing or decreasing. We also demonstrated the application of multiple low-cost sensors in evaluating indoor air quality during an extreme wildfire smoke episode. We suggest that long-term use of low-cost sensors can aid facility operators in testing and optimizing actions aimed at protecting occupants from wildfire smoke infiltration. These findings could inform the development of building guidelines by local health authorities for both new construction and renovation of healthcare facilities. As wildfire seasons become longer and more intense, understanding smoke infiltration in healthcare facilities is important to reduce indoor exposure for more susceptible populations.

Author Contributions

Conceptualization, G.M., G.E., K.J., N.Z. and S.B.H.; Formal analysis, P.D.M.N. and N.M.; Methodology, P.D.M.N., N.M., N.Z. and S.B.H.; Supervision, N.Z. and S.B.H.; Visualization, P.D.M.N. and N.M.; Writing—original draft, P.D.M.N. and N.M.; Writing—review and editing, G.M., G.E., K.J., N.Z. and S.B.H. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Vancouver Coastal Health Authority (Internal Project # D049-00) and Health Canada (MOA # 4500415138).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are not publicly available. However, most data can be visualized in an interactive Shiny web application: https://ehs-bccdc.shinyapps.io/facility_AQdata/. The “Concentrations” tab shows time series of calibrated data of indoor locations after indoor peaks removal along with the calibrated rooftop and raw data from FEM stations. The “Ratio” tab shows the immediate indoor–outdoor ratio during the study period between all indoor locations and the rooftop, fitted with a smoothing spline.

Acknowledgments

We gratefully thank the staff at the facility for supporting the installation of the air quality sensors and providing insights into the facility operations. We also gratefully thank colleagues at Metro Vancouver for providing air quality and meteorological data used in the analyses.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Ford, B.; Martin, M.V.; Zelasky, S.E.; Fischer, E.V.; Anenberg, S.C.; Heald, C.L.; Pierce, J.R. Future Fire Impacts on Smoke Concentrations, Visibility, and Health in the Contiguous United States. GeoHealth 2018, 2, 229–247. [Google Scholar] [CrossRef] [Green Version]
  2. Liu, J.C.; Pereira, G.; Uhl, S.A.; Bravo, M.A.; Bell, M.L. A systematic review of the physical health impacts from non-occupational exposure to wildfire smoke. Environ. Res. 2014, 136, 120–132. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  3. Reid, C.; Brauer, M.; Johnston, F.H.; Jerrett, M.; Balmes, J.R.; Elliott, C.T. Critical Review of Health Impacts of Wildfire Smoke Exposure. Environ. Health Perspect. 2016, 124, 1334–1343. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  4. Zheng, X.-Y.; Ding, H.; Jiang, L.-N.; Chen, S.-W.; Zheng, J.-P.; Qiu, M.; Zhou, Y.-X.; Chen, Q.; Guan, W.-J. Association between Air Pollutants and Asthma Emergency Room Visits and Hospital Admissions in Time Series Studies: A Systematic Review and Meta-Analysis. PLoS ONE 2015, 10, e0138146. [Google Scholar] [CrossRef]
  5. Yee, J.; Cho, Y.A.; Yoo, H.J.; Yun, H.; Gwak, H.S. Short-term exposure to air pollution and hospital admission for pneumonia: A systematic review and meta-analysis. Environ. Health 2021, 20, 1–10. [Google Scholar] [CrossRef] [PubMed]
  6. Dominici, F.; Peng, R.D.; Bell, M.; Pham, L.; McDermott, A.; Zeger, S.L.; Samet, J.M. Fine Particulate Air Pollution and Hospital Admission for Cardiovascular and Respiratory Diseases. JAMA 2006, 295, 1127–1134. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  7. Matz, C.J.; Egyed, M.; Xi, G.; Racine, J.; Pavlovic, R.; Rittmaster, R.; Henderson, S.B.; Stieb, D.M. Health impact analysis of PM2.5 from wildfire smoke in Canada (2013–2015, 2017–2018). Sci. Total Environ. 2020, 725, 138506. [Google Scholar] [CrossRef]
  8. Leech, J.A.; Nelson, W.C.; Burnett, R.T.; Aaron, S.; Raizenne, M.E. It’s about time: A comparison of Canadian and American time–activity patterns. J. Expo. Sci. Environ. Epidemiol. 2002, 12, 427–432. [Google Scholar] [CrossRef] [Green Version]
  9. Flannigan, M.; Cantin, A.S.; de Groot, W.J.; Wotton, M.; Newbery, A.; Gowman, L.M. Global wildland fire season severity in the 21st century. For. Ecol. Manag. 2013, 294, 54–61. [Google Scholar] [CrossRef]
  10. Davison, G.; Barkjohn, K.K.; Hagler, G.S.W.; Holder, A.L.; Coefield, S.; Noonan, C.; Hassett-Sipple, B. Creating Clean Air Spaces During Wildland Fire Smoke Episodes: Web Summit Summary. Front. Public Health 2021, 9, 508971. [Google Scholar] [CrossRef] [PubMed]
  11. Wettstein, Z.S.; Hoshiko, S.; Fahimi, J.; Harrison, R.J.; Cascio, W.E.; Rappold, A.G. Cardiovascular and Cerebrovascular Emergency Department Visits Associated with Wildfire Smoke Exposure in California in 2015. J. Am. Heart Assoc. 2018, 7, e007492. [Google Scholar] [CrossRef] [Green Version]
  12. Pantelic, J.; Dawe, M.; Licina, D. Use of IoT sensing and occupant surveys for determining the resilience of buildings to forest fire generated PM2.5. PLoS ONE 2019, 14, e0223136. [Google Scholar] [CrossRef]
  13. Barn, P.; Larson, T.I.; Noullett, M.; Kennedy, S.; Copes, R.; Brauer, M. Infiltration of forest fire and residential wood smoke: An evaluation of air cleaner effectiveness. J. Expo. Sci. Environ. Epidemiol. 2007, 18, 503–511. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  14. Mousavi, E.S.; Kananizadeh, N.; Martinello, R.A.; Sherman, J.D. COVID-19 Outbreak and Hospital Air Quality: A Systematic Review of Evidence on Air Filtration and Recirculation. Environ. Sci. Technol. 2020, 55, 4134–4147. [Google Scholar] [CrossRef]
  15. Rajagopalan, P.; Goodman, N. Improving the Indoor Air Quality of Residential Buildings during Bushfire Smoke Events. Climate 2021, 9, 32. [Google Scholar] [CrossRef]
  16. Reisen, F.; Powell, J.C.; Dennekamp, M.; Johnston, F.H.; Wheeler, A. Is remaining indoors an effective way of reducing exposure to fine particulate matter during biomass burning events? J. Air Waste Manag. Assoc. 2019, 69, 611–622. [Google Scholar] [CrossRef] [PubMed]
  17. Zhao, X.; Sun, Y.; Zhao, C.; Jiang, H. Impact of Precipitation with Different Intensity on PM2.5 over Typical Regions of China. Atmosphere 2020, 11, 906. [Google Scholar] [CrossRef]
  18. Chithra, V.; Nagendra, S.S. Impact of outdoor meteorology on indoor PM10, PM2.5 and PM1 concentrations in a naturally ventilated classroom. Urban Clim. 2014, 10, 77–91. [Google Scholar] [CrossRef]
  19. Xu, R.; Qi, X.; Dai, G.; Lin, H.; Zhai, P.; Zhu, C.; Wang, L.; Ding, A. A Comparison Study of Indoor and Outdoor Air Quality in Nanjing, China. Aerosol Air Qual. Res. 2020, 20, 1–14. [Google Scholar] [CrossRef] [Green Version]
  20. Wan, Y.; Chen, C.; Wang, P.; Wang, Y.; Chen, Z.; Zhao, L. Infiltration Characteristic of Outdoor Fine Particulate Matter (PM2.5) for the Window Gaps. Procedia Eng. 2015, 121, 191–198. [Google Scholar] [CrossRef] [Green Version]
  21. Liang, D.; Lee, W.-C.; Liao, J.; Lawrence, J.; Wolfson, J.M.; Ebelt, S.T.; Kang, C.-M.; Koutrakis, P.; Sarnat, J.A. Estimating climate change-related impacts on outdoor air pollution infiltration. Environ. Res. 2021, 196, 110923. [Google Scholar] [CrossRef] [PubMed]
  22. Braniš, M.; Řezáčová, P.; Domasová, M. The effect of outdoor air and indoor human activity on mass concentrations of PM10, PM2.5, and PM1 in a classroom. Environ. Res. 2005, 99, 143–149. [Google Scholar] [CrossRef] [PubMed]
  23. Taylor, J.; Mavrogianni, A.; Davies, M.; Das, P.; Shrubsole, C.; Biddulph, P.; Oikonomou, E. Understanding and mitigating overheating and indoor PM2.5 risks using coupled temperature and indoor air quality models. Build. Serv. Eng. Res. Technol. 2015, 36, 275–289. [Google Scholar] [CrossRef]
  24. Giordano, M.R.; Malings, C.; Pandis, S.N.; Presto, A.A.; McNeill, V.; Westervelt, D.M.; Beekmann, M.; Subramanian, R. From low-cost sensors to high-quality data: A summary of challenges and best practices for effectively calibrating low-cost particulate matter mass sensors. J. Aerosol Sci. 2021, 158, 105833. [Google Scholar] [CrossRef]
  25. Holder, A.L.; Mebust, A.K.; Maghran, L.A.; McGown, M.R.; Stewart, K.E.; Vallano, D.M.; Elleman, R.A.; Baker, K.R. Field Evaluation of Low-Cost Particulate Matter Sensors for Measuring Wildfire Smoke. Sensors 2020, 20, 4796. [Google Scholar] [CrossRef]
  26. Delp, W.W.; Singer, B.C. Wildfire Smoke Adjustment Factors for Low-Cost and Professional PM2.5 Monitors with Optical Sensors. Sensors 2020, 20, 3683. [Google Scholar] [CrossRef]
  27. Gupta, P.; Doraiswamy, P.; Levy, R.; Pikelnaya, O.; Maibach, J.; Feenstra, B.; Polidori, A.; Kiros, F.; Mills, K.C. Impact of California Fires on Local and Regional Air Quality: The Role of a Low-Cost Sensor Network and Satellite Observations. GeoHealth 2018, 2, 172–181. [Google Scholar] [CrossRef]
  28. Wicked Device LLC. Air Quality Egg [Internet]; Wicked Device LLC: Ithaca, NY, USA, 2018; Available online: https://airqualityegg.com/egg (accessed on 22 March 2021).
  29. R Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2020; Available online: https://www.R-project.org/ (accessed on 14 January 2021).
  30. Carslaw, D.C.; Ropkins, K. openair—An R package for air quality data analysis. Environ. Model. Softw. 2012, 27–28, 52–61. [Google Scholar] [CrossRef]
  31. Malings, C.; Tanzer, R.; Hauryliuk, A.; Saha, P.K.; Robinson, A.L.; Presto, A.A.; Subramanian, R. Fine particle mass monitoring with low-cost sensors: Corrections and long-term performance evaluation. Aerosol Sci. Technol. 2019, 54, 160–174. [Google Scholar] [CrossRef]
  32. Sonntag, D. Important New Values of the Physical Constants of 1986, Vapour Pressure Formulations Based on the ITS-90, and Psychrometer Formulae. Zeitschrift für Meteorologie 1990, 70, 340–344. [Google Scholar]
  33. Liland, K.H.; Almøy, T.; Mevik, B.-H. Optimal Choice of Baseline Correction for Multivariate Calibration of Spectra. Appl. Spectrosc. 2010, 64, 1007–1016. [Google Scholar] [CrossRef]
  34. Allen, R.; Larson, T.; Sheppard, L.; Wallace, L.; Liu, L.-J.S. Use of Real-Time Light Scattering Data to Estimate the Contribution of Infiltrated and Indoor-Generated Particles to Indoor Air. Environ. Sci. Technol. 2003, 37, 3484–3492. [Google Scholar] [CrossRef] [PubMed]
  35. Baltagi, B.H. (Ed.) Distributed Lags and Dynamic Models. In Econometrics; Springer: Berlin/Heidelberg, Germany, 2008; pp. 129–145. [Google Scholar] [CrossRef]
  36. Gasparrini, A. Distributed lag linear and non-linear models in R: The package dlnm. J. Stat. Softw. 2011, 43, 1–20. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  37. Wheeler, A.; Allen, R.; Lawrence, K.; Roulston, C.; Powell, J.; Williamson, G.; Jones, P.; Reisen, F.; Morgan, G.; Johnston, F. Can Public Spaces Effectively Be Used as Cleaner Indoor Air Shelters during Extreme Smoke Events? Int. J. Environ. Res. Public Health 2021, 18, 4085. [Google Scholar] [CrossRef] [PubMed]
  38. Xiang, J.; Huang, C.-H.; Shirai, J.; Liu, Y.; Carmona, N.; Zuidema, C.; Austin, E.; Gould, T.; Larson, T.; Seto, E. Field measurements of PM2.5 infiltration factor and portable air cleaner effectiveness during wildfire episodes in US residences. Sci. Total Environ. 2021, 773, 145642. [Google Scholar] [CrossRef] [PubMed]
  39. Yao, J.; Stieb, D.M.; Taylor, E.; Henderson, S.B. Assessment of the Air Quality Health Index (AQHI) and four alternate AQHI-Plus amendments for wildfire seasons in British Columbia. Can. J. Public Health 2019, 111, 96–106. [Google Scholar] [CrossRef]
  40. May, N.W.; Dixon, C.; Jaffe, D.A. Impact of Wildfire Smoke Events on Indoor Air Quality and Evaluation of a Low-cost Filtration Method. Aerosol Air Qual. Res. 2021, 21, 210046. [Google Scholar] [CrossRef]
  41. Canadian Standards Association. Special Requirements for Heating, Ventilation, and Air-Conditioning (HVAC) Systems in Health Care Facilities (CSA Standard Z317.2:19) [Internet]. 2019. Available online: https://www.csagroup.org/store/product/CSA%20Z317.2%3A19/ (accessed on 14 July 2021).
  42. Cichowicz, R.A.; Dobrzański, M. Indoor and Outdoor Concentrations of Particulate Matter (PM10, PM2.5) and Gaseous Pollutants (VOC, H2S) on Different Floors of a University Building: A Case Study. J. Ecol. Eng. 2021, 22, 162–173. [Google Scholar] [CrossRef]
  43. Krebs, B.; Burney, J.; Zivin, J.G.; Neidell, M. Using Crowd-Sourced Data to Assess the Temporal and Spatial Relationship between Indoor and Outdoor Particulate Matter. Environ. Sci. Technol. 2021, 55, 6107–6115. [Google Scholar] [CrossRef]
  44. ASHRAE. Planning Framework for Protecting Commercial Building Occupants from Smoke during Wildfire Events [Internet]. 2021. Available online: https://www.ashrae.org/file%20library/technical%20resources/covid-19/guidance-for-commercial-building-occupants-from-smoke-during-wildfire-events.pdf (accessed on 1 August 2021).
Figure 1. Floor plan of the facility with locations of low-cost sensors (red dots).
Figure 1. Floor plan of the facility with locations of low-cost sensors (red dots).
Ijerph 18 09811 g001
Figure 2. Locations of the rehabilitation facility (red dot) and three National Air Pollution Surveillance (NAPS) monitoring stations (blue dots) at Vancouver International Airport (YVR), Clark Drive, and Kensington Park in greater Vancouver, Canada. Location of greater Vancouver relative to North America (red square on inset map).
Figure 2. Locations of the rehabilitation facility (red dot) and three National Air Pollution Surveillance (NAPS) monitoring stations (blue dots) at Vancouver International Airport (YVR), Clark Drive, and Kensington Park in greater Vancouver, Canada. Location of greater Vancouver relative to North America (red square on inset map).
Ijerph 18 09811 g002
Figure 3. Time series of raw 1-min PM2.5 concentrations from the low-cost sensors located outdoors on the rehabilitation facility rooftop and from beta attenuation monitors at three national air pollution surveillance (NAPS) monitoring stations at Kensington Park, Vancouver International Airport (YVR) and Clark Drive in greater Vancouver, Canada (Figure 2).
Figure 3. Time series of raw 1-min PM2.5 concentrations from the low-cost sensors located outdoors on the rehabilitation facility rooftop and from beta attenuation monitors at three national air pollution surveillance (NAPS) monitoring stations at Kensington Park, Vancouver International Airport (YVR) and Clark Drive in greater Vancouver, Canada (Figure 2).
Ijerph 18 09811 g003
Figure 4. Time series of calibrated 1-min PM2.5 concentrations from the low-cost sensors located outdoors (blue) and indoors (grey). Indoor peaks have not been removed from these data. The cause of the short peak on October 15 is unknown.
Figure 4. Time series of calibrated 1-min PM2.5 concentrations from the low-cost sensors located outdoors (blue) and indoors (grey). Indoor peaks have not been removed from these data. The cause of the short peak on October 15 is unknown.
Ijerph 18 09811 g004
Figure 5. Heat maps of 2-h averaged PM2.5 concentrations at the rehabilitation facility during the wildfire smoke episode. The low cost sensor on the rooftop is shown at the top, and those located indoors are shown below. The y-axis of each plot indicates the time of day. Indoor peaks have not been removed from these data. Data from the morning of September 11 show that smoke moved rapidly indoors as the rooftop concentration started to increase, and overnight data from September 18–19 show that smoke cleared rapidly out of the building as rooftop concentration decreased.
Figure 5. Heat maps of 2-h averaged PM2.5 concentrations at the rehabilitation facility during the wildfire smoke episode. The low cost sensor on the rooftop is shown at the top, and those located indoors are shown below. The y-axis of each plot indicates the time of day. Indoor peaks have not been removed from these data. Data from the morning of September 11 show that smoke moved rapidly indoors as the rooftop concentration started to increase, and overnight data from September 18–19 show that smoke cleared rapidly out of the building as rooftop concentration decreased.
Ijerph 18 09811 g005
Table 1. The 1-min mean PM2.5 concentration, the Pearson r, the mean absolute error (MAE), and the coefficient of variation of the mean absolute error (CvMAE) of the six project Air Quality Egg (EGG) lost-cost sensors that were co-located at Kensington Park (Figure 2) prior to the study period. The EGG 4 and EGG 6 sensors were later deployed to the facility inside the ground floor lobby and outside a second-floor window, respectively. However, EGG 6 failed shortly after the beginning of the study period.
Table 1. The 1-min mean PM2.5 concentration, the Pearson r, the mean absolute error (MAE), and the coefficient of variation of the mean absolute error (CvMAE) of the six project Air Quality Egg (EGG) lost-cost sensors that were co-located at Kensington Park (Figure 2) prior to the study period. The EGG 4 and EGG 6 sensors were later deployed to the facility inside the ground floor lobby and outside a second-floor window, respectively. However, EGG 6 failed shortly after the beginning of the study period.
Mean PM2.5
(µg/m3)
Pearson rMean MAE
(µg/m3)
Mean CvMAE
EGG 15.920.822.700.50
EGG 213.770.734.880.91
EGG 35.820.832.610.48
EGG 47.210.783.010.56
EGG 56.950.832.660.49
EGG 66.210.822.730.51
Table 2. Infiltration coefficients and their relative ranks for indoor locations in the rehabilitation center after removing indoor peaks during the smoke episode and typical days.
Table 2. Infiltration coefficients and their relative ranks for indoor locations in the rehabilitation center after removing indoor peaks during the smoke episode and typical days.
LocationNumber of Indoor Peaks RemovedInfiltration CoefficientRank
Typical DaysSmoke EpisodeTypical DaysSmoke Episode
Fourth floor patient room1380.380.3723
Fourth floor office30.260.3157
Fourth floor elevator60.330.3544
Ground floor lobby140.360.4532
Ground floor therapy room1590.220.3176
Basement office80.270.3465
Basement elevator640.390.4711
Table 3. The 24-h mean and range of PM2.5 concentrations during the smoke episode and typical days. Vancouver International Airport (YVR), Clark Drive, and Kensington Park are ambient air quality monitoring stations in the National Air Pollution Surveillance (NAPS) network (Figure 2). At the rehabilitation facility, the rooftop sensor was outdoors while the other seven were indoors. Completeness of datasets is shown as percentages out of 103,680 expected 1-min data points from 00:00 21 August to 23:59 31 October.
Table 3. The 24-h mean and range of PM2.5 concentrations during the smoke episode and typical days. Vancouver International Airport (YVR), Clark Drive, and Kensington Park are ambient air quality monitoring stations in the National Air Pollution Surveillance (NAPS) network (Figure 2). At the rehabilitation facility, the rooftop sensor was outdoors while the other seven were indoors. Completeness of datasets is shown as percentages out of 103,680 expected 1-min data points from 00:00 21 August to 23:59 31 October.
LocationSmoke Episode Typical DaysData Completeness (%)
24-h Mean (µg/m3)Range (µg/m3)24-h Mean (µg/m3)Range (µg/m3)
Greater Vancouver
YVR69.19.5–144.94.50.9–14.396.7
Clark Drive75.811.1–161.47.11.5–46.296.1
Kensington Park76.17.7–172.44.81.0–17.697.2
The facility
Rooftop72.07.7–141.67.50.0–46.862.5
Fourth floor patient room31.82.9–55.13.50.2–14.298.4
Fourth floor office23.91.2–44.61.10.0–8.699.8
Fourth floor elevator28.32.0–50.21.60.0–11.098.6
Ground floor lobby34.01.5–63.42.00.1–13.597.0
Ground floor therapy room24.54.0–45.53.40.3–19.399.1
Basement office28.42.6–52.51.90.0–11.299.8
Basement elevator36.52.6–71.83.00.0–16.299.2
Table 4. The 24-h average of meteorological conditions during the smoke episode and typical days. All indoor sensors were aggregated in the indoor facility calculations. Outdoor sensors are the rooftop air quality egg, Clark Drive, Kensington Park, and Vancouver International Airport (YVR) which are part of the National Air Pollution Surveillance (NAPS) network.
Table 4. The 24-h average of meteorological conditions during the smoke episode and typical days. All indoor sensors were aggregated in the indoor facility calculations. Outdoor sensors are the rooftop air quality egg, Clark Drive, Kensington Park, and Vancouver International Airport (YVR) which are part of the National Air Pollution Surveillance (NAPS) network.
Location and ParameterSmoke EpisodeTypical Days
Facility indoors
Temperature (°C)22.221.9
Relative Humidity (%)48.344.4
Facility rooftop
Temperature (°C)17.915.8
Relative Humidity (%)66.069.8
Clark Drive
Temperature (°C)17.711.3
Relative Humidity (%)77.584.4
Wind Speed (km/h)2.74.0
Wind Direction (Degree)153.3150.1
Precipitation (mm)0.00.0
Kensington Park
Temperature (°C)17.811.0
Relative Humidity (%)76.184.9
Wind Speed (km/h)5.77.0
Wind Direction (Degree)149.6132.2
Precipitation (mm)0.00.0
YVR
Temperature (°C)16.811.0
Relative Humidity (%)80.182.5
Wind Speed (km/h)7.810.0
Wind Direction (Degree)165.4160.6
Precipitation (mm)0.00.0
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Nguyen, P.D.M.; Martinussen, N.; Mallach, G.; Ebrahimi, G.; Jones, K.; Zimmerman, N.; Henderson, S.B. Using Low-Cost Sensors to Assess Fine Particulate Matter Infiltration (PM2.5) during a Wildfire Smoke Episode at a Large Inpatient Healthcare Facility. Int. J. Environ. Res. Public Health 2021, 18, 9811. https://0-doi-org.brum.beds.ac.uk/10.3390/ijerph18189811

AMA Style

Nguyen PDM, Martinussen N, Mallach G, Ebrahimi G, Jones K, Zimmerman N, Henderson SB. Using Low-Cost Sensors to Assess Fine Particulate Matter Infiltration (PM2.5) during a Wildfire Smoke Episode at a Large Inpatient Healthcare Facility. International Journal of Environmental Research and Public Health. 2021; 18(18):9811. https://0-doi-org.brum.beds.ac.uk/10.3390/ijerph18189811

Chicago/Turabian Style

Nguyen, Phuong D. M., Nika Martinussen, Gary Mallach, Ghazal Ebrahimi, Kori Jones, Naomi Zimmerman, and Sarah B. Henderson. 2021. "Using Low-Cost Sensors to Assess Fine Particulate Matter Infiltration (PM2.5) during a Wildfire Smoke Episode at a Large Inpatient Healthcare Facility" International Journal of Environmental Research and Public Health 18, no. 18: 9811. https://0-doi-org.brum.beds.ac.uk/10.3390/ijerph18189811

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop