On the Applicability of Low-cost Sensors for Measurements of Aerosol Concentrations

Abstract

One of the most important parameters of air pollution, concentration of particulate matter suspended in the air (aerosols), is measured by means of low-cost sensors of latest generation. Such measurements have high resolution in both space and time, but there is a big question about applicability of low-cost sensors for measurements of aerosol concentrations that arise from the complexity of measurement technique, and inevitable errors associated. In this paper we analyze all aspects of applicability of low-cost sensors for long-term measurement campaigns.

Full text

30TH DAAAM INTERNATIONAL SYMPOSIUM ON INTELLIGENT MANUFACTURING AND AUTOMATION
DOI: 10.2507/30th.daaam.proceedings.060
ON THE APPLICABILITY OF LOW-COST SENSORS FOR
MEASUREMENTS OF AEROSOL CONCENTRATIONS
Adnan Masic, Dzevad Bibic & Boran Pikula
This Publication has to be referred as: Masic, A[dnan]; Bibic, D[zevad] & Pikula, B[oran] (2019). On the Applicability
of Low-cost Sensors for Measurements of Aerosol Concentrations, Proceedings of the 30th DAAAM International
Symposium, pp.0452-0456, B. Katalinic (Ed.), Published by DAAAM International, ISBN 978-3-902734-22-8, ISSN
1726-9679, Vienna, Austria
DOI: 10.2507/30th.daaam.proceedings.060
Abstract
One of the most important parameters of air pollution, concentration of particulate matter suspended in the air (aerosols),
is measured by means of low-cost sensors of latest generation. Such measurements have high resolution in both space
and time, but there is a big question about applicability of low-cost sensors for measurements of aerosol concentrations
that arise from the complexity of measurement technique, and inevitable errors associated. In this paper we analyze all
aspects of applicability of low-cost sensors for long-term measurement campaigns.
Keywords: Low-cost sensor; Air pollution; Aerosol concentration; Calibration.
1. Introduction
Air pollution is a major issue nowadays: it affects many aspects of our lives, from the health (directly) to the economy
(indirectly). There are various components of air pollution, and among all, particles suspended in the air (aerosols) are
primary pollutant in most cases [1], [2]. In order to assess the level of air pollution by aerosols, we need to measure their
concentrations in the air. But not all aerosols are equally harmful. Particles are classified by their size, and generally,
smaller particles are more dangerous. There are many standards and recommendations for particle concentrations,
proposed from different organizations. All of them are based on the concentrations of particles smaller than 10 m (PM10)
or 2.5m (PM2.5). In this study, we will use PM2.5 measurements, since they appear in most studies (PM10 is slowly
becoming deprecated for this purpose). The problem is the fact that proper measurement of aerosol concentrations is
complex and expensive. Thus, for most polluted areas (poor and developing countries), we don’t have much (if any) data.
Recently, new technology has emerged in this field: the low-cost sensors. Table 1 shows main classification of devices
for PM2.5 concentration measurements. Most accurate methods are based on gravimetric measurements, but they are
expensive and the resolution is very limited in both space and time. Equivalent methods (usually based on the attenuation
of beta radiation, BAM) are not as accurate as referent methods, but they are cheaper to run and offer better time
resolution. Research grade portable devices, usually based on optical measurement techniques, have even lower cost,
better resolution, but they may give inaccurate results, due to the principle of their work. Low-cost sensors are very
popular because of their price, size and time resolution (typically 1 s) but the question is: can they be used for
measurements of PM concentrations at all? We will try to answer this question in the present paper.
- 0452 -

30TH DAAAM INTERNATIONAL SYMPOSIUM ON INTELLIGENT MANUFACTURING AND AUTOMATION
Method
Reference
Equivalent
Research grade calibrated
Low-cost
Device
example
Cost
Very high
High
Moderate
Very low
Accuracy
Best
Good
Moderate/good, depending
on conditions
Questionable
Typ. time
resolution
1 day
1 hour
1 second
1 second
Portability
Very low
Low
High
Very High
Table 1. Classification of devices for measurement of aerosol concentrations
There are many papers in which performance of low-cost aerosol sensors is evaluated, for example [3], [4], [5]. However,
articles with long campaigns in realistic conditions are very rare. In this paper results of successful long-term comparison
of low-cost sensor with professional BAM monitor nearby are presented.
2. Low-cost sensor of choice
Taking into account conclusions from our previous studies [5], [6] among all commercially available low-cost sensors,
we have selected Plantower PMS5003 sensor. Main reasons for such choice are reliability of the sensors, consistence
between units from different batches of sensors, well implemented digital interface and good mean time before failure
(MTBF). It is optical based sensor which uses Mie scattering theory for laser scattering on spherical particles to calculate
size and concentration of particles in the unit volume of air. This approach, however, inherits several shortcomings for
accurate measurement of mass concentrations of aerosols, including:
- assumption that particles are spherical,
- inability to detect particles smaller than 0.3 m in diameter,
- assumption for index of refraction of particles,
- assumption of particles’ density.
These parameters may change in the atmosphere over time, which means that there is no correct and permanent factory
calibration. Furthermore, high relative humidity in the air may lead to the effect of hygroscopic growth of aerosols [7],
[8], which leads to overshoots in measurements at such conditions.
Low-cost sensors are not weather proof, which means that we need to provide proper enclosure. That is not an easy
task, because the enclosure must not obstruct sampling of aerosols. Putting PMS5003 sensor into simple box with some
openings for the air is not solution, because the sensor has both inlet and outlet on the same side, and such enclosure
would lead to inaccurate results due to the cyclone effect. Data acquisition must be performed flawlessly because
measurement loses are not acceptable here. All these challenges were previously solved within the framework of our in-
house developed MAQS sensor ecosystem and data acquisition system [9].
3. Case study
The city of Sarajevo is affected by aerosol air pollution very much, and it was selected as our case study. Sarajevo is
situated in the natural valley, with strong episodes of temperature inversions during winter [10], [11] and very low number
of windy days. Results of this study may be relevant for other cities with similar topology of terrain as well. Figure 1
shows the locations of devices for the measurement campaign. We set up two MAQS sensors: MAQS@MEF was located
at the Faculty of Mechanical Engineering while MAQS@ETF was installed at the Faculty of Electrical Engineering,
University of Sarajevo. The referent device was professional BAM monitor (equivalent method in Table 1) located inside
of the complex of US Embassy in Sarajevo (BAM@Embassy). Data collected from the BAM@Embassy are publicly
available, but access to the Embassy device was not possible. Strictly speaking, two MAQS sensors were not collocated
with BAM, but they were close to each other (200 m from ETF to Embassy and 500 m from MEF to Embassy), so that it
is unlikely that atmosphere can differ significantly at such short distances.
- 0453 -

30TH DAAAM INTERNATIONAL SYMPOSIUM ON INTELLIGENT MANUFACTURING AND AUTOMATION
Fig. 1. Location of measurement devices
4. Results
Data from Embassy monitor were collected from 23/MAY/2018, MAQS@MEF started data acquisition on 7/17/2018
and MAQS@ETF started on 2/7/2019. All devices are running continuously as of today, 10/18/2019, which is the last
day included in this paper. During this period, many different scenarios were recorded.
Figure 2 shows comparisons of daily average PM2.5 values between two MAQS sensors and BAM@Embassy. The
coefficient of determination between MAQS and BAM was sensational: 𝑅2= 0.96 for MAQS@MEF and 𝑅2= 0.93 for
MAQS@ETF (please note that MAQS@MEF collected more data than MAQS@ETF, which started after the peak of air
pollution during winter 2018/19). The absolute values, however, didn’t agree well: MAQS@MEF measured 37% higher
value (on average) than BAM@Embassy, while MAQS@ETF recorded 31% higher value than BAM (on average).
Fig. 2. Daily average values of PM2.5 - comparisons between MAQS sensors and BAM
Figure 3 shows comparisons of hourly average values. Different factors associated to the principle of work of optical
aerosol sensor appear as noise in measurements on such timescale. The fact that sensors are not strictly collocated may
also play some role here. BAM monitor is not perfectly accurate either, so it is natural, that correlation is weaker in case
of hourly average values. But the results are still good: correlation between MAQS sensors and BAM device is high: 𝑅2=
0.86 for MAQS@MEF and 𝑅2= 0.79 for MAQS@ETF. Absolute values, again, are overestimated by similar factors as
in the case of hourly averages.
y = 0.7295x - 2.0544
R² = 0.9567
0
50
100
150
200
250
300
350
400
450
0100 200 300 400 500
BAM@Embassy
MAQS@MEF
PM2.5 daily avg from 7/17/18 [g/m3]
y = 0.7368x + 0.1248
R² = 0.9312
0
20
40
60
80
100
120
050 100 150
BAM@Embassy
MAQS@ETF
PM2.5 daily avg from 2/7/19 [g/m3]
- 0454 -

30TH DAAAM INTERNATIONAL SYMPOSIUM ON INTELLIGENT MANUFACTURING AND AUTOMATION
Fig. 3. Hourly average values of PM2.5 - comparisons between MAQS sensors and BAM
Consistency between two MAQS sensors is also important. Figure 4 shows daily and hourly average values for two
MAQS sensors. Since the sensors are not collocated (distance between them was about 300 m) we don’t expect perfect
match, but they should be well correlated. Indeed, daily averages had correlation coefficient 𝑅2= 0.98 while hourly
averages had 𝑅2= 0.96, which is excellent. Their absolute values were close (well within 10%).
Fig. 4. Daily and hourly average values of PM2.5 - comparisons between MAQS sensors at MEF and ETF
5. Conclusion
Low-cost sensors have potential for monitoring of aerosol concentrations: they are affordable and robust (at least those
used in this study), they have modern digital interface and can be adapted for outdoor use with proper enclosure. However,
low-cost sensors can’t be used out-of-the-box for direct measurements of mass concentrations. They must be calibrated,
which may be difficult and expensive. If there is any reference or equivalent public station nearby, that is excellent
opportunity for comparison and (some kind of) calibration of low-cost sensors. In such cases low-cost sensors could fill
the gaps between professional stations, taking into account their time and space resolution.
As suggestion for future work we propose the following activities:
- investigation of humidity influence on low-cost sensors and development of model of correction,
- measurement of concentrations of nanoparticles, which can’t be detected with low-cost sensors,
- strict colocation between low-cost sensors and reference monitor,
- using reference (gravimetric) methods for calibration of low-cost sensors under various conditions.
y = 0.6963x - 0.7673
R² = 0.8642
-100
0
100
200
300
400
500
600
0200 400 600
BAM@Embassy
MAQS@MEF
PM2.5 hourly avg from 7/17/18 [g/m3]
y = 0.6756x + 1.2625
R² = 0.793
0
20
40
60
80
100
120
140
160
180
200
050 100 150 200 250
BAM@Embassy
MAQS@ETF
PM2.5 hourly avg from 2/7/19 [g/m3]
y = 0.944x + 0.7886
R² = 0.9826
0
20
40
60
80
100
120
050 100 150
MAQS@ETF
MAQS@MEF
PM2.5 daily avg from 2/7/19 [g/m3]
y = 0.9047x + 2.6842
R² = 0.9577
0
50
100
150
200
250
300
350
400
0100 200 300 400
MAQS@ETF
MAQS@MEF
PM2.5 hourly avg from 2/7/19 [g/m3]
- 0455 -

30TH DAAAM INTERNATIONAL SYMPOSIUM ON INTELLIGENT MANUFACTURING AND AUTOMATION
5. Acknowledgments
We would like to thank to Mr. Damir Muslić for his generous contribution to our project by writing core part of the in-
house developed MAQS software.
6. References
[1] https://www.epa.gov/pm-pollution/health-and-environmental-effects-particulate-matter-pm, (2019). United States
Environmental Protection Agency, Accessed on: 2019-10-18
[2] https://www.who.int/airpollution/data/cities/en, (2018). WHO Global Ambient Air Quality Database (update 2018),
Accessed on: 2019-10-18
[3] Crilley, L. R.; Shaw, M.; Pound, R.; Kramer, L. J.; Price, R.; Young, S.; Lewis, A. C. & Pope F. D.(2018).
Evaluation of a Low-cost Optical Particle Counter (Alphasense OPC-N2) for Ambient Air Monitoring. Atmospheric
Measurement Techniques, 11, 709–720, 2018, DOI: https://doi.org/10.5194/amt-11-709-2018
[4] Chatzidiakou, L. et al. Characterising low-cost sensors in highly portable platforms to quantify personal exposure
in diverse environments. Atmospheric Measurement Techniques, 12, 4643–4657, 2019, DOI:
https://doi.org/10.5194/amt-12-4643-2019
[5] Masic, A.; Pikula, B. & Bibic, D. (2017). Mobile Measurements of Particulate Matter Concentrations in Urban Area,
Proceedings of the 28th DAAAM International Symposium, pp.0452-0456, B. Katalinic (Ed.), Published by
DAAAM International, ISBN 978-3-902734-11-2, ISSN 1726-9679, Vienna, Austria DOI:
10.2507/28th.daaam.proceedings.063
[6] Masic, A.; Pikula, B.; Bibic, Dz.; Musemic, R. & Halac, A. (2018). Calibration and Assessment of Low-cost Dust
sensors, Proceedings of the 29th DAAAM International Symposium, pp.0523-0528, B. Katalinic (Ed.), Published
by DAAAM International, ISBN 978-3-902734-20-4, ISSN 1726-9679, Vienna, Austria, DOI:
10.2507/29th.daaam.proceedings.075
[7] Chen, J. et al. Aerosol hygroscopic growth, contributing factors, and impact on haze events in a severely polluted
region in northern China. Atmospheric Chemistry and Physics, 19, 1327–1342, 2019, DOI:
https://doi.org/10.5194/acp-19-1327-2019
[8] Jayaratne, R.; Liu, X.; Thai, P.; Dunbabin, M. & Morawska, L. (2018). The Influence of Humidity on the
Performance of a Low-cost Air Particle Mass Sensor and the Effect of Atmospheric Fog. Atmospheric Measurement
Techniques, Atmospheric Measurement Techniques, 11, 4883–4890, DOI: https://doi.org/10.5194/amt-11-4883-
2018
[9] Masic, A. (2015). Unmanned Aerial Vehicle as Data Acquisition System. Journal of Trends in the Development of
Machinery and Associated Technology. Vol. 19, No. 1, 2015, ISSN 2303-4009 (online), p.p. 181-184.
[10] Masic, A.; Musemic, R. & Dzaferovic-Masic, E. (2016). Temperature Inversion Measurements in Sarajevo Valley
Using Unmanned Aerial Vehicles, Proceedings of the 27th DAAAM International Symposium, pp.0423-0427, B.
Katalinic (Ed.), Published by DAAAM International, ISBN 978-3-902734-08-2, ISSN 1726-9679, Vienna, Austria,
DOI: 10.2507/27th.daaam.proceedings.062
[11] Masic, A.; Bibic, Dz.; Pikula, B.; Dzaferovic-Masic, E. & Musemic, R. Experimental study of temperature
inversions above urban area using unmanned aerial vehicle. Thermal Science, 2018, DOI:
10.2298/TSCI180227250M
- 0456 -