Air Quality Measurements in Kitchener, Ontario, Canada Using Multisensor Mini Monitoring Stations
The findings presented above on the quantitative metrics for performance highlight that: (1) for Oand PM2.5, the multisensor pods are an acceptable and practical alternative to the more expensive research grade equipment used in reference stations. For NO, our analysis suggests that LOC of 10 ppb reported by the manufacturer needs revision particularly from low pollution sites like Kitchener, (2) there is room for improvement in sensor technology for these pollutants, particularly for NOand PM2.5, and (3) it is important to assess the performance of multisensor pods in the field using research grade equipment. The stationary nature of the MECP monitoring stations hampers the performance assessment of the other four sensors near the schools. In an attempt to circumvent this issue, a custom-made long distance scaling tool was developed. This tool allowed for further calibration of the collected data from the datasets obtained from the AQMesh multisensor pods using US EPA certified calibration techniques [ 24 ]. In addition, this analysis tool enabled the isolation of outlier data points caused by varying emission sources across the five sites. A more comprehensive explanation of this tool can be found in the Supplementary document
Data points for the levels of PM2.5 in Figure 2 C,F show that the majority are clustered below 15 μg m, the WHO AQG for PM2.5, in the fall ( Figure 2 C) and winter ( Figure 2 F) seasons. During the summer season ( Figure 2 I), higher variability in PM2.5 levels was recorded, impacted by wildfires in Northern Ontario. There was a relatively high correlation between the sensor PM2.5 data and that from the reference station with R> 0.7 for the fall and winter, which drops to R~ 0.6 in the summer.
3.2. Impacts of Pollutant Emissions on the Local Air Quality
25,2, O3, and PM2.5 concentrations to rank the quality of air on a scale of 1–10, with one indicating low risk to health and ten indicating a high risk to health. Equation (1) is used for the calculation of the AQHI:
AQHI
=
(
1000
10.4
)
x
[
(
e
0.000537
·
[
O
3
]
−
1
)
+
(
e
0.000871
·
[
NO
2
]
−
1
)
+
(
e
0.000487
·
[
PM
2.5
]
−
1
)
]
(1)
The Air Quality Health Index (AQHI) is a metric used in Ontario to assess air quality from data measured at the reference stations across the province [ 4 26 ]. This index incorporates a rolling 3-h average of NO, O, and PM2.5 concentrations to rank the quality of air on a scale of 1–10, with one indicating low risk to health and ten indicating a high risk to health. Equation (1) is used for the calculation of the AQHI:
2 and O3 in the calculation of the AQHI [
The derivation of Equation (1) relied on the number of deaths recorded across ten major cities in Canada as a result of an elevated pollutant, as opposed to adverse effects experienced by both healthy and vulnerable communities with prolonged exposure to said pollutant. In addition, Equation (1) is an outdated formula with the last revision taking place over a decade ago [ 27 28 ]. The MECP uses a variation of Equation (1) termed the AQHI+. This variation involves the inclusion of the Air Quality Index (mAQI) for NOand Oin the calculation of the AQHI [ 29 ]. This additional parameter only takes effect when either of these pollutants reach a relatively high concentration. In these scenarios, If the mAQI value exceeds the calculated AQHI value and is larger than six, then the former is classified as the final AQHI+ value. A flow chart highlighting the procedure followed to calculate the AQHI+ can be found in the Supplementary document (Figure S1) . To conserve the consistency in this study, this parameter was applied to our datasets across the multisensor pod network.
An analysis of PM2.5 data from the sensor pods indicated that readings are inaccurate during high relative humidity events (i.e., rainy days). This is due to a multitude of factors including, but not limited to: “swelling” of particulates, which causes larger particulate size measurements by the optical particle counter, deliquescence build-up on sensor intake yielding inaccurate readings, and water droplets entering the inlet and classified as particulates. To circumvent this issue, wet days, which are defined as days with rain, hail, or snow in any volume, were omitted from further analysis with the AQHI+ equation. In addition, personal communications with the MECP scientists provided crucial feedback on the criteria for data quality used to calculate the AQHI+. Hence, the following conditions were applied to the AQMesh data sets:
-
If precipitation was observed for a particular day, PM2.5 was omitted from the AQHI calculations,
-
If data were available for only one pollutant, the AQHI was recorded as 1, and,
-
If the AQHI was calculated to be zero, then the AQHI was recorded as 1.
2 levels, as detailed in 2, to assess the effect of NO2 on the sensitivity of the AQHI values. The results from these sensitivity assessments showed that NO2 concentrations were dominant in the morning hours (7–10 a.m.), evident by the shift in the peak frequency from an AQHI value of 2 with the inclusion of NO2, to an AQHI value of 1 with the removal of NO2. Using the same testing parameters for the afternoon hours (3–6 p.m.), O3 was found to be the dominant species during pickup times. This is evident by the absence of a peak frequency shift in the AQHI values with and without NO2 removal.
Using the calculations and conditions stated above, histograms were constructed for the month of October for dry workdays (see Supplementary Figure S2 ). These histograms show the number of times (i.e., frequency) the AQHI value was at a certain number for both school drop off (7–10 a.m.) and pickup (3–6 p.m.) times to determine the dominant pollutant contributing to the AQHI during these intervals. The analysis was done using data from the reference station, the co-located site (Pod 5), and a third arbitrarily selected site (Pod 2). Due to the under-performance of the sensor pod in measuring NOlevels, as detailed in Section 3.1 , two types of AQHI calculations were done, with and without NO, to assess the effect of NOon the sensitivity of the AQHI values. The results from these sensitivity assessments showed that NOconcentrations were dominant in the morning hours (7–10 a.m.), evident by the shift in the peak frequency from an AQHI value of 2 with the inclusion of NO, to an AQHI value of 1 with the removal of NO. Using the same testing parameters for the afternoon hours (3–6 p.m.), Owas found to be the dominant species during pickup times. This is evident by the absence of a peak frequency shift in the AQHI values with and without NOremoval.
To assess the variations of AQHI+ across the pod network and in relation to the reference station, histograms were generated for each time period studied for both drop off and pickup times. These histograms compared the frequency of AQHI+ values recorded at each sensor pod in the AQMesh network in addition to comparisons with the reference station. These results, shown in Figure 3 , illustrate that the multisensor network detects different levels of pollutants at each location, which can be attributed to the different emission sources near each pod location.
32,
In addition, the impacts of seasonal variations on air quality are also observable in these figures. Figure 3 A–C highlight the drop off time comparisons for fall, winter, and summer seasons, respectively. The peak frequency for the reference station is located at the AQHI+ value of 2 for all time periods, while the pod network is observed to have a lower frequency at this value caused by a higher frequency of AQHI+ values exceeding 2. This is most apparent in Figure 3 C, where there are distinguishable peaks in the frequency of AQHI+ values ranging from 3 to 7. The analysis for the pickup times in Figure 3 D–F highlights the increase in the AQHI+ value across all sites during pickup times. This observation is most apparent in Figure 3 F, where the peak frequency for each sensor pod is larger than the value reported by the MECP. Pod 1, located near a highway, including access routes to the highway, has a peak at an AQHI+ value of 5, which is higher than the value reported by the MECP. Neighborhood scale variations in pollution are well known to be found in larger metropolitan cities [ 30 ]. The results in Figure 3 highlight that these variations can be found in smaller, less polluted cities, which is further supported by the findings of other research groups [ 31 33 ]. This variability in ambient contaminants was more precisely studied by Richards et al. [ 34 ], where a mobile mass spectrometer was used to map out the volatile organic compound sources across eastern Vancouver Island. These findings underline the inability of a single reference station to detect local hotspot regions in Kitchener, much less across the Region of Waterloo, and exemplifies the benefits of monitoring air quality using a multisensor network.
Percent
mismatch
=
N
u
m
b
e
r
o
f
d
a
y
s
p
o
d
w
a
s
d
i
f
f
e
r
e
n
t
t
h
a
n
r
e
f
e
r
e
n
c
e
s
t
a
t
i
o
n
T
o
t
a
l
n
u
m
b
e
r
o
f
d
a
y
s
×
100
%
(2)
To further assess the AQHI+ values calculated for each pod in the network against the reference station, percentage mismatch comparisons were made for each period. This simple mismatch calculation per Equation (2) calculates the percentage of individual AQHI+ data points that exceed the reported values by the MECP for the same point in time for both drop off and pickup intervals by a index value of 1 or greater:
2, the AQHI+ values are also dependent on the availability of light [Figure 4 A shows the percentage mismatch comparisons for the drop off time interval. This figure highlights the difference in pollutant emissions at each pod location. Pod 1, located near the highway, shows the largest mismatch values for all three seasons studied. This finding further supports the previous conclusion that the AQHI+ values are strongly influenced by traffic counts and local emission sources. Furthermore, this finding indicates that the current provincial monitoring station is not sufficient in accurately detecting AQHI+ fluctuations at the neighborhood scale, highlighting the need for pod networks to identify pollutant hotspots and the cause of these hotspots. When comparing the seasonal mismatch percentages in Figure 4 A, it becomes apparent that the mismatch is also impacted by seasonal variations as mentioned in the long-distance scaling Section 3.3 . With the formation of secondary pollutants being dependent on the availability of light to photochemically dissociate NO, the AQHI+ values are also dependent on the availability of light [ 22 35 ]. Percentage mismatch values during the winter periods in Figure 4 are lower than the three seasons studied as they might have been influenced by the lockdown restrictions put in place between 23 December 2020 to 26 January 2021. Lingering restrictions over the following six weeks as shown in reference [ 36 ] and Supplementary Table S1 might also have contributed to this trend. This lockdown severely impacted the daily traffic counts which have reduced emissions during this period [ 23 37 ].
