approximates that by the year 2020, there will be

Mobile Air Quality around 50 billion things connected to the internet [2].

Monitoring with Edge This will surely create bottlenecks in future, the signs
of which are becoming visible even now. Edge
Computing computing and Fog computing are the new solutions
that complement, but do not completely replace cloud
computing. The purpose of these approaches is to
bring computing power closer to the devices that
generate data. This heralds the rise of a new set of
Abstract:
applications that were impossible to implement
With the oncoming wave of IoT applications and big
hitherto, while also improving the efficiency of older
companies taking steps to create Edge computing
systems. This is especially true for IoT devices and
platforms, the world is taking another leap forward in
related systems as many of these systems generate
technological advancement. However, can these
large amounts of unfiltered data and their need for
advancements help improve the current approaches
time sensitive decision making.
to monitoring air quality? This paper discusses the
benefits of Edge computing and how it can be used for However, the new paradigms do not come without
a more effective Air Quality Monitoring System. difficulties of their own. While the benefits of these
approaches are enormous, so is the complexity of
their implementation. Presently, a large number of
I. Introduction companies, which created cloud computing platforms
for the development and deployment of IoT solutions,
are attempting to provide edge computing platforms,
We’ve seen a boom in innovations and solutions in the
while Microsoft has specifically launched Azure IoT
last two decades. One of the most notable ones being
Edge which is a hybrid Cloud-Edge computing service
the advent of Cloud computing, an architecture that
[3].
has enabled so many of the applications we use today.
However, even though Cloud computing is an This paper discusses edge computing in the context of
architecture that only gained significant traction in air quality monitoring, especially addressing its
late 1990’s [1], it has quickly become an unfit solution application in polluted cities in Pakistan along with a
for today’s problems. Cloud technologies involve few general applications. The structure of the paper is
significant latency and are not suitable for time as follows:
sensitive applications. Furthermore the cloud
paradigm has been constrained by data privacy issues ● A discussion about the edge computing paradigm
such as in health monitoring and home devices. and its different approaches, followed by its
Therefore, the solutions that Cloud computing definition, and the advantages it has over
approach could provide are becoming limited in the traditional cloud computing with the help of use
backdrop of growing needs of the present. As the case examples and a discussion of problems and
needs of today evolve, technology also must evolve challenges that arise with adopting the edge
with it. Different approaches, namely fog computing computing architecture.
and edge computing, have thus since arisen, which ● Overview of air quality monitoring and the need
address the various shortcomings of cloud computing for it. This section contains sample statistics of
such as inefficient usage of network bandwidth, heavily populated cities of Pakistan.
unnecessary communication between devices and the ● Use case of edge computing in the context of air
cloud and reliance on a single point of failure. quality monitoring to demonstrate its benefits in
the context of the chosen use case.
With the Internet of Things (IoT) technologies and use
cases growing exponentially, it won’t be long before
the vast amount of data produced saturates our II. Overview of Edge Computing
networks. Cisco Internet Business Solutions Group
At the outset, it must be stated that admittedly edge received from every node to make future predictions.
computing cannot be defined clearly and precisely, as Another benefit of offloading processing data to the
yet. edge is the decreased latency that the application will
experience.
The architecture is relatively vague in the sense that
there is no exact documentation [4] defining how its Low latency [4] has become a requirement for many
architecture should be implemented. Different bodies applications and this cannot be achieved with cloud
have proposed different ways to approach the computing as all the computing is carried out at a
architecture. Some suggest creating small data centers distant location with transmission delays in between,
close to the edge access networks that will provide resulting in slower and less desirable user
little to no delay and give a rich real-time experience experiences.
to application users. Another approach known as
mobile edge computing (MEC) combines IT and B. Bandwidth and Privacy
networking, which integrates server capabilities at the With the number of IoT devices increasing
edge of cellular networks allowing for many benefits exponentially, the huge amount of data generated by
such as relief of core network congestion, better them will be a waste of bandwidth usage, since the
application hosting services and increased real-time cloud rarely needs all of the data generated. In Edge
response to applications such as augmented reality. A computing, nodes will filter through the data using
more common and simpler way to look at edge application-based algorithms and only send data that
computing is to have node devices such as sensors at is actually needed by the cloud. This on one hand will
the edge of the network that generate data, and do decrease the workload for the cloud, which would
part of the computing. lead to marginally faster response time. But more
importantly, the bandwidth used to send this filtered
This would mean, as one paper puts it, that data data through the network will be significantly more
producers would become data consumers [6][5] and than that needed for blindly sending unfiltered data,
their computing and storage power would be which will result in less network usage costs as well.
leveraged for a quicker response and decreased load
on the cloud. There are many ways to approach the Privacy has been a huge problem for cloud computing
same concept but we would like to formally define the applications as not all data collected, such as smart
unchanging idea all these different methods employ. home or health monitoring data, should be sent over
to the cloud for analysis as this violates the privacy of
That is, Edge computing is the placement of application users. Edge computing solves this by
computing power closer to the source of data filtering through data and only sending data that is
generation. To further explain what Edge computing is, void of privacy issues and can cache private data at
a few reasons will be highlighted that explain why the edge and consume it locally thus protecting user
there was need for it and how it has a clear advantage data from ever leaving their local devices.
over the traditional cloud approach.
C. Applications
A. Computing Offloading Having briefly discussed how edge computing
As scalable as cloud computing is, the devices reliant overcomes the problems of cloud computing, here are
upon the cloud for computational and storage services a few popular practical examples of edge computing
will keep increasing. At some point, the relatively easy applications that benefit from the previously
scalability of the node devices creates a certain discussed edge computing improvements.
amount of “unnecessary” load for the cloud. With all
these devices trying to access and use the cloud, the 1) Augmented Reality
services to each of these devices become limited and We are now seeing a phenomenal increase in the
are constrained. This load on the cloud can be popularity of augmented reality applications. Just
substantially reduced [6] if the devices themselves looking at the success of the Ingress made augmented
perform part of the computational tasks at the edge reality game [4] Pokemon GO makes it obvious that
and cached data while requesting services from the there is going to be a suite of similar such applications
cloud only for complex functions and decision making available in the near future. These applications
such as learning patterns from a large amount of data depend on the physical location of the devices and
thus have a core need for low latency and quick Working with the edge architecture makes
response timing as the real world location influences programming much more complex than if it were just
the application and has to be modified and depicted on the cloud. Working with differing runtimes
accurately within a desirable interval to make user between differing edge nodes is an additional
experience smoother. Similarly, all applications that challenge for programmers. Coupled with is the
attempt to create an immersive experience like problem of dynamically naming edge devices [6].
multi-user video games benefit from low latency. DNS and URI naming systems might not be optimal for
the slew of edge devices to be employed
2) Smart Homes [6]
Smart home devices are plentiful in the market such
as smart security systems, smart energy systems and III. Air Quality Control
even smart temperature control systems. Other than
these complex systems, there are numerous smaller
devices that work towards making a home smart. All This section discusses the issue of increasingly
these devices usually need to constantly collect user worsening air quality and why it is of utmost urgency
influenced data to create the best experience. As most to take steps to monitor it accurately and in real-time
of these systems collect valuable personal data of to potentially save not only lives but economies as
homeowners, many privacy related promises have to well.
be maintained when collecting user data, which cloud
The quality of air we breathe is one of the key
computing cannot guarantee. On the other hand, Edge
determinants of the quality of our lives. [8] However,
computing enables users to consume much of the
unfortunately, only 1 in 10 people breathe safe and
data at the edge and send only suitably filtered data to
clean air worldwide, according to WHO, whereas in
the cloud. Edge computing architecture allows private
56% of the cities and towns being monitored, the
data to be stored locally with selective access.
average air pollution level is 3.5 times higher than the
D. Problems WHO recommended safe air quality standard. Also,
With all the apparent benefits, there are many 87% of the deaths from air pollution happen in low
challenges that arise with the adoption of the Edge and middle income countries.
computing paradigm. A few of those problems are:
Air pollution is the cause of many deaths and diseases
Offloading strategies: With an increased number of around the world. It is a major issue and many
devices to manage, efficient strategies need to be governments are proactively working to create
employed to handle the collaboration between legislations and air quality standards that should be
multiple devices. Even with the processing power maintained globally and locally to avoid air pollution
enabled at the edge, the benefits of this approach from becoming more of a health hazard than it
cannot be fully reaped without optimal workload already is. Common results of increased air pollution
allocation [6] [7]. As this approach has many layers are Acid rain, Eutrophication, [9] Ozone and
with different processing, storage and analytical respiratory and heart problems [10].
capabilities, edge computing implementations need to
Sample statistics for the disease related effects are as
optimize their capabilities, as each layer is better
follows [11]:
suited to process data in a certain situation. However,
implementing this intelligence is easier said than done ● 25% of all deaths and disease from lung
as many factors play a part in making these decisions cancer
such as size of data, priority of data and network ● 17% of all deaths and disease from acute
bandwidth etc. lower respiratory infection
● 16% of all deaths from stroke 
Complexity and Cost: The increased amount of overall ● 15% of all deaths and disease from ischemic
computational power does not come without costs of heart disease
its own. Installing newer functionality inside edge
devices will also increase the cost of the edge devices.
Well known hazardous air pollutants include Ozone
gas created by chemical reactions at surface level,
Sulfur Dioxide from combustion of fossil fuels, and also a contributor to emissions that add into
Particulate Matter (PM) formed by similar means. Particulate Matter and also cause Ozone layer
Particulate Matter in particular is a very dangerous depletion.
health hazard and research by the World Health
Organization indicates that once absorbed into the B. Statistics in Pakistan
bloodstream, it can have cardiovascular, Taking a closer look at Pakistan, which is one of the
cerebrovascular and respiratory impacts. In 2013, it largest developing third world countries, being the
was classified as a cause of lung cancer by WHO’s sixth most populous country in the word, there are
International Agency for Research on Cancer (IARC) many alarming statistics reported by the World Health
[11]. There are two types of PM - PM10 and PM2.5 - Organization. [8] Pakistan is reportedly the 4th most
that differ in diameter, which is the most commonly polluted country in the world and air pollution
used measurement of air quality. [8] Between, 2008 mappings across Pakistan shows that the air quality
and 2013, pollution levels around the globe increased index in many areas is way above the standard limits.
by 8%, higher than any other time period in the
Here are some statistics of the most densely
history of the world.
populated cities in Pakistan.
A. Causes of Air pollution [10]
1. Karachi

[8] The largest city of Pakistan with a population of

1. Burning of Fossil Fuels: 
almost 20 million has a PM2.5 measurement of 88
As economies grow and people’s livelihoods grow with µg/m3, which is 8.8 times above the WHO safe level.
them, there is a greater reliance on fossil fuels. These
fuels turn the wheels of the industry and are also
burnt in most of the transportation vehicles. The more
they are emitted into the atmosphere after
combustion, the more pollution they create and their
overuse is catastrophic for the environment. Common
fossil fuel combustion byproducts include dangerous
gases like Sulfur Dioxide, Carbon monoxide and
Nitrogen Oxides.
2. Lahore
2. Agricultural activities: 
[8] Lahore has an annual average of 68 µg/m3,
Supplementary chemicals used in agriculture release which is 6.8 times the WHO safe level.
Ammonia gas as a byproduct which is one of the most
hazardous gases in the atmosphere. Overuse of
pesticides and fertilizer has resulted in harmful
compounds that affect both air and water.

3. Mining operations:

During mining operations, processes that extract

minerals from the earth, the resulting dust generated
and chemicals released from the Earth’s surface
account for a significant part of the air pollution 3. Rawalpindi
problem. This is the main reason for the worsening
[8] At 107 µg/m3, the average annual air pollution
health of workers in mines and nearby residents.
in Rawalpindi is surprisingly even higher than
4. Indoor air pollution: Karachi and Lahore and 10.7 times the WHO safe
level.
 The household products we use every day to either
keep our houses clean or make them smell fresh are
 and although the technology employed in these
facilities is supposedly great at what they do and can
produce very accurate results, it is limited by being
tied down to a few fixed locations. This means that
even though the reports from dedicated stations may
be highly accurate, they cannot be referred to for
locations, which are distant from these installations.
These stations can collect data, merge it, analyze it
and give an overall estimate for the entire area they
These are only average statistics meaning that the are supposed to cover. However, for health and
peak values are unfortunately much higher than agriculture related issues more accurate and real time
what is shown in these reports. data is needed for smaller spaces and specific
locations. As these stations employ large and
expensive technology, they are limited in number
IV. Edge Computing for Air Quality which results in “spatial gaps” [1] meaning the data
collected and analyzed by them does not accurately
portray the air quality conditions of sparsely
Air quality monitoring is definitely a government level
populated localities.
responsibility since it has such a large scale impact and
in fact millions of dollars are spent on the research Governments usually have legislative requirements to
and improvement of air quality monitoring techniques fulfill [14] and so must install monitoring networks in
by governments of developed countries and even such a way that they concentrate on collecting data
some developing countries. The main aim of this from areas with a higher population density. Thus, it is
paper is to advocate for the implementation of Air difficult to draw conclusions about how air quality
Quality monitoring through the Edge computing varies from place to place, if you only have a few data
paradigm. points [14]. This is especially true for rural areas [1]
where stations are distributed even more sparsely.
A. WHY?
Why should air quality monitoring be done any 2) Temporal Gaps
differently than the current approach? This section Aside from this, the fact that air quality monitoring
discusses the reasons why the adopted approach to and environmental analysis is expensive means that
air quality monitoring is not optimal as a standalone not only the stations are geographically limited but
solution. Firstly, what is the commonly adopted they also cannot provide real-time services for all the
approach? different types of pollutant measurements to save on
cost. This practice results in “temporal gaps” [1]
Countries may allocate a budget that will be based on meaning that some types of emissions will only be
the recent needs to monitor air quality in an area, only sampled after set periods of time, that could be even
after the negative effects of bad air quality have many days. Unfortunately, air quality does not remain
significantly affected the area. In the case of third stable enough to justify timely gaps between samples
world countries this budget is used to purchase and this means air quality data is not very accurate in
machinery and equipment from developed countries. real time.
For instance, [9] Pakistan spent 80 million of its
currency to purchase ambient air quality monitoring Air quality may fluctuate for a number of reasons [15].
systems from Germany in the year 2016-2017. Most commonly due to excessive pollution thrown
Another part of the budget is spent on the training into the air at a certain moment in time. A large event
involved in operating and using the advanced such as a wildfire can dramatically darken the skies.
technologies. Due to the high energy requirements on hotter days,
fuel consumption increases significantly, leading to
1) Spatial Gaps higher than normal levels of emission. Sometimes
The problem is that these technologies are likely to local culture may also be a reason for high energy
work as connected stations across an area of interest.
Currently, money is spent on labs, stations and towers
requirements. Research shows a trend of higher measure and report air quality to cloud-based
energy consumption in the US on weekends. platforms in near real time. In this context the possible
solution would be to use public transport such as taxis
Weather conditions can also influence the amount of and buses to mount air quality sensors and monitor
air pollution in the atmosphere. For example, sunlight air pollution levels. As the vehicles move city wide
is needed in the chemical reactions that create ozone throughout the day, this would help to identify poor
gas [15]. Inversion [15] between warm and cold air air quality hotspots. The collected data which is
may result in pollutants building up and increasing in accumulated at the cloud can be parsed by machine
concentration in a certain space as they have nowhere learning algorithms and either work together with
to escape to. This is common in mountainous areas. data from stations or simply be used as a secondary
mechanism to monitor air quality.
Sometimes pollution may suddenly arrive from
entirely different places of origin. Emissions from
somewhere else in the world may be carried over to a
different area to cause a sudden rise in air pollutants.

B. HOW?
Seeing the limitations of the traditional fixed station
approach, it is easy to see that an extension is
required to fill in the gaps seen in localized air quality
data.

One solution that has gained much traction over

recent years is “Mobile Air Quality Monitoring”
(MAQM) [16]. This section talks about the advantages The benefits of MAQM are clear as the network of
of implementing MAQM in a large area by using Edge sensors through an area will significantly increase the
computing technologies. data collected and thus result in more accurate
reports about different zones and can highlight which
Developing third world countries typically face serious zones are more critical needing drastic remedial
budget constraints and purchasing large sensing and measures, as well aiding in informing the public to
analysis machinery means that in the future, a lot of avoid such areas. This will not only help to fill the
money will be needed for the maintenance of these geographical and temporal gaps, but will also provide
machines, which is a huge challenge for these more accurate data as it will accumulate a lot more
countries. With the world moving toward a more samples over time. However, all this data collected
connected era and sensors for IoT devices becoming from hundreds and thousands of sensors is going to
more common and less costly while coming in a wide put a strain on the network bandwidth and might
variety of calibrations, cheaper bandwidth even overshadow the cost of purchase and installation
consumption and availability of low cost air quality of such systems in the long run. To avoid simply
sensors, these can be used to complement the data of collecting data from all types of sensors and solely
fixed stations by providing more real time and denser relying on a cloud based platform to compute through
data that will more accurately indicate pollutant levels all the acquired data, it is necessary to employ Edge
in specific locations than overall average computed by computing strategies.
reference stations [14]. Although there are various
types of air pollution sensors, and some are
specialized in certain aspects, the majority focuses on
five components: ozone, particulate matter, carbon
monoxide, sulfur dioxide, and nitrous oxide. Now that
we live in a period in which these sensors are more
portable, cheap, easy to make and easy to deploy, a
mesh of mobile sensors can be used to determine the
real state of air quality within an area for potentially
better handling of health hazards. These sensors can
 to many isolated incidents such as fires and
explosions.

Another benefit from having smarter edge nodes

would be to change calibrations on the go and not
worry about recalibrating every sensor that is
deployed. Controlling sensors through the cloud could
grant automatic calibration for active devices in the
field.

C. Problems and Limitations

Using an IoT solution for air quality monitoring has a

few limitations as well.

1) Calibrations
Even with cheaper sensors available, due to hardware
limitations, it is not possible for these sensors to give
the level of accuracy and work with the same
Edge computing for mobile air quality monitoring is a calibrations provided by expensive technology found
solution that incorporates all the benefits of Edge in stations. Due to less accuracy as compared to
computing platforms for an air quality monitoring stations, these sensors can only give indicative
system. It uses a connected mesh of air quality measures of air quality such as percentages [17]
sensors deployed in the field, either installed into rather than fine grain small measurements.
vehicles or at specific locations of interest, that act as 2) Height
edge nodes having their own dedicated processing Sensors that are very close to the ground level will not
units and memory buffers. be able to collect data at a height. Air pollution is not
With a mix of Edge computing and Cloud computing, evenly distributed vertically and so this might cause
these low cost sensors gradually become smarter in discrepancies in samples collected versus the actual
their data collection and analysis. Small detached state of the atmosphere.
sensors take in many samples of data every minute 3) Cost
but not all this data needs to be sent back to the cloud Since edge computing enabled sensors will only be
as many times data can give an inaccurate adding to an already constrained budget, the benefits
representation as it may have outliers that do not of saving on cost will not be so quickly realized as they
correctly reflect the current atmosphere. Instead a will only take effect in the long run.
running average is a much more efficient piece of data
that could be used by the cloud for further
calculations. V. Conclusion
This article has talked about the rapid growth and
As sensors are supposed to be fitted with global
benefit of Edge computing. It then moved on to the
positioning systems (GPS) to send location information
dangers of air pollution and the current state of
constantly so that air quality data can be associated
developing third world countries. Then it lists why the
with different zones. Also sensors can be made smart
current air quality monitoring approaches may not be
enough to not send the data when they are not on the
a complete and accurate assessment of the
move or have not significantly changed their location.
atmosphere and suggests how a mobile network of
In many such ways, the redundant data can be filtered
edge nodes can fill in the gaps of the current system.
out, so that the backend calculations work much
faster. This will result in quick responses to the real
time air quality conditions that can be mercurial due
 [11]. World Health Organization, Ambient air
pollution: Health impacts,

Reference Page http://www.who.int/airpollution/ambient/hea

lth-impacts/en/, Accessed on March 29th.
[12].Conserve Energy Future, © 2018,
[1]. CEHTP, Air Quality: Measurements and their https://www.conserve-energy-future.com/
limitations, causes-effects-solutions-of-air-pollution.ph
http://www.cehtp.org/faq/air/air_quality_me p, Accessed on March 29th.
asures_and_limitations, Accessed on March
29th.
[2]. Dave Evans, The Internet of Things How the
Next Evolution of the Internet Is Changing [13]. Rameez Khan, Govt to install five ambient
Everything, April 2011, air quality monitoring systems, June 5,
https://www.cisco.com/c/dam/en_us/about/a 2017,
c79/docs/innov/IoT_IBSG_0411FINAL.pdf, https://tribune.com.pk/story/1427466/govt
Accessed on April 29th. -install-five-ambient-air-quality-monitoring-s
[3]. Linknovate Team, Edge Computing Leaders ystems/, Accessed on March 29th.
you Must Know, February 14, 2018, [14]. Nuria Castell, Franck R. Dauge, Philipp
https://blog.linknovate.com/edge-computing-l Schneider, Matthias Vogt, Uri Lerner b ,
eaders-you-must-know/, Accessed on March Barak Fishbain , David Broday , Alena
29th. Bartonova, Can commercial low-cost sensor
[4]. Gopika Premsankar, Mario Di Francesco, and platforms contribute to air quality
Tarik TalebEdge Computing for the Internet of monitoring and exposure estimates?, 28
Things: A Case Study, Accessed on April 29th. December
[5]. JIALE ZHANG, BING CHEN, YANCHAO ZHAO, 2016,https://ac.els-cdn.com/S01604120163
XIANG CHENG, AND FENG HU, Data Security 09989/1-s2.0-S0160412016309989-main.pd
and Privacy-Preserving in Edge Computing, 28 f?_tid=be339d2b-f280-4386-9bc9-f2aa8d5e
March 2018, Accessed on March 29th. 2e0b&acdnat=1524914631_645826a172db
[6]. Weisong Shi, Jie Cao, Quan Zhang, Youhuizi Li, 31928bda15b954149147, Accessed on
and Lanyu Xu, 5th October 2016, Edge March 29th.
Computing: Vision and Challenges, Accessed [15]. Holli Riebeek, Robert Simmon, Why Monitor
online: 2nd March 2018. Air Quality, May 19, 2009,
[7]. Blesson Varghese, Nan Wang, Sakil Barbhuiya, https://terra.nasa.gov/citizen-science/air-qu
Peter Kilpatrick and Dimitrios S. Nikolopoulos, ality, Accessed on March 29th.
Challenges and Opportunities in Edge [16]. Jason Thongplang, The benefits of mobile
Computing, 7 Sep 2016, Accessed on April air quality monitoring, Jan 15, 2016,
29th. https://www.aeroqual.com/mobile-air-quali
[8]. Bilal Hakim, Gasping for Air: Here’s How Major ty-monitoring, Accessed on March 29th.
Pakistani Cities Rank for Air Pollution, Jan 24th [17]. Amanda Randle, The challenges and
2017, benefits of local air quality monitoring, Oct
https://propakistani.pk/2017/01/24/gaspi 05 2016,
ng-air-heres-major-pakistani-cities-rank-air https://www.envirotech-online.com/article/
-pollution/, Accessed on March 29th. air-monitoring/6/aqmesh/the-challenges-an
[9]. Ecschooltoday, d-benefits-of-local-air-quality-monitoring/21
http://eschooltoday.com/pollution/air-poll 04, Accessed on March 29th.
ution/effects-of-air-pollution.html, Accessed
on March 29th.
[10].Rinkesh,
https://www.conserve-energy-future.com/
causes-effects-solutions-of-air-pollution.ph
p, Accessed on March 29th.