IEEE Consumer Electronics Magazine - January/February 2023 - 28

Toward Blue Skies
Table 2. Mean and range of pollutant concentrations for the residential (RES) and industrial
(IND) districts.
Pollutant
PM1:0
RES
PM2:5
IND
PM10
PM1:0
PM2:5
PM10
Mean
2:04 1:59 mg=m3
2:07 1:66 mg=m3
2:13 1:63 mg=m3
31:45 22:83 mg=m3
1136:77 1404:99 mg=m3
20633:55 28378:82 mg=m3
Range (min-max)
0:71mg=m3-6:25mg=m3
0:72mg=m3-6:42mg=m3
0:72mg=m3- 6:42mg=m3
2:50mg=m3- 71:19mg=m3
2:55mg=m3-3798:15mg=m3
2:55mg=m3-75391:60mg=m3
main source of high particle concentrations, it
also captures and binds other pollutants. At
higher altitudes, the pollutants either fall to the
ground or are dispersed in the environment,
depending on whether the pollutants are
released below or above the so-called inversion
layer. The altitude at which inversion occurs
varies as a function of temperature, the shape of
the landscape, and other factors.16 Thus, monitoring
the vertical column and capturing pollutants
at different altitudes is essential for
capturing the dispersion patterns of pollutants
and understanding the factors that affect it.
Indeed, monitoring the vertical column also
increases the coverage of pollutant information
and enables a detailed view of the pollutant distribution
and the mechanisms governing its dispersion
in an urban environment.
Correlation and Significance Testing
For the residential site, Spearman's r
between PM concentration and altitude shows a
strong negative and statistically significant correlation
for all three particle sizes: PM1:0, r ¼
0:7, p< 0:05; PM2:5, r ¼0:6, p< 0:05; and
PM10, r ¼0:7, p< 0:05. For the industrial site,
the direction of the correlations is reversed as the
r values are positive and statistically highly significant:
PM1:0, r ¼ 0:8, p< 0:01; PM2:5, r ¼ 0:9,
p< 0:01;and PM10, r ¼ 0:9, p< 0:01.These
results thus further demonstrate that the
differences in vertical distribution are both significant
and dependent on the sampling location.
Understanding the full extent of health
risks and the processes that govern the dispersion
of pollutants requires capturing these
variations.
28
Combined Effect of Altitude and Location
The full potential of UAVs comes their use to
capture the 3D distribution of pollutants. Next, we
assess the joint effect that location and altitude
have on the measurements by considering each
10-meter increase in altitude as a separate interval
and simultaneously comparing the differences in
pollution distributions across the two locations
and the different intervals. In line with earlier analysis,
a repeated measures ANOVA shows significant
main effects for location and altitude
interval, but no interaction effect (location: F ¼
10:048, p< 0:01, h2 ¼ 0:167; altitude interval: F ¼
10:059, p< 0:01, h2 ¼ 0:167; interaction: F ¼
1:788, p ¼ 0:094, h2 ¼ 0:243). Here, Frefers to theFstatistic
of the ANOVA, and h2 is the proportion of
variance that is explained. Posthoc comparisons
(Bonferroni correction) indicate statistically significant
differences from70-meters onward. Differences
in emissions can thus be at highly different
altitudes and monitoring both horizontal and vertical
dimensions is required to capture the full
extent of pollutants.
Comparison to Reference Station Baseline
The final part of the experiments contrasts
the UAV measurements with those provided by
reference stations. The reference sensors are
particle matter measurements provided by the
Environmental Services Unit of the City of Kotka.
The city has three measurement stations, and
we matched our measurements with those of the
closest station. The reference stations are
located close to ground level, but the altitude of
the sensing units varies from approximately 4-8meters
depending on the reference station.
The comparison uses the altitude of the
IEEE Consumer Electronics Magazine

IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023