This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

Design and Implementation of LPWA-based Air Quality

Monitoring System

Kan Zheng∗ , Senior Member, IEEE, Shaohang Zhao∗ , Zhe Yang∗ , Xiong Xiong∗ , Wei Xiang† , Senior
Member, IEEE
∗
Intelligent Computing and Communication (IC2 ) Lab
Key Lab of Universal Wireless Communications, Ministry of Education
Beijing University of Posts & Telecommunications
Beijing, China, 100088
†
College of Science Technology & Engineering Division of Tropical Environments and Societies
James Cook University Cairns
QLD 4870 Australia
Contact email: zkan@bupt.edu.cn

Abstract

Increasing attention has been paid to air quality monitoring with a rapid development in industry and
transportation applications in the modern society. However, existing air quality monitoring systems cannot
provide satisfactory spatial and temporal resolutions of the air quality information with low costs in real-
time. In this paper, we propose a new method to implement the air quality monitoring system based on
state-of-the-art Internet-of-Things (IoT) techniques. In this system, portable sensors collect the air quality
information timely, which is transmitted through a low power wide area (LPWA) network. All air quality
data are processed and analyzed in the IoT cloud. The completed air quality monitoring system including
both hardware and software is developed and deployed successfully in urban environments. Experimental
results show that the proposed system is reliable in sensing the air quality, which helps reveal the change
patterns of air quality to some extent.

Index Terms– Internet-of-Things (IoT), air quality monitoring, PM2.5 , LPWA, IEEE 802.15.4k

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

I. I NTRODUCTION

With the rapid development in industry and transportation, air pollution has recently become a serious
problem for developing countries, which has been paid increased attention to by both governments and

the public. It is well-accepted that the people exposed to air pollutants for long periods of time are more
likely to suffer from severe respiratory diseases [1]. If air quality continues to deteriorate, the cost of
pollution harness may become a heavy burden for governments. Thus, air quality monitoring systems are

very useful in effectively monitoring air pollution before the situation becomes worse.

Traditionally, air quality monitoring stations are usually of large sizes and high costs for installation

and maintenance, which limits its potential in dense deployment in cities [2]. Furthermore, although the

precise measurement results can be produced, time-consuming procedures are needed offline. Thus, air

quality data cannot be provided in real time by this way. However, air quality information data of both

high spatial and temporal resolution in both the spatial and temporal dimensions is highly desired, which

is the focal point of this paper [3].

Thanks to the rapid development of Internet-of-Things (IoT) technology, the air quality can be sensed

and the corresponding data transmitted to the servers through wireless networks such as the wireless

sensor networks (WSNs) [4] [5] [6] [7]. Various portable air quality sensors with acceptable costs have

been used. One of the main air pollution sources is the tiny particles with an aerodynamic diameter of less

than 2.5 micrometer, i.e., Particulate Matter (PM) 2.5, which can attach various poisonous chemicals and

be breathed into human lungs. Thus, the PM sensor is a good choice for providing reliable measurement

data on the air dust level in cities [8] [9]. Without a complicated process, the sensor can measure the
air quality in seconds. Although the accuracy of these sensors might not be as good as traditional air

quality monitoring devices, it is fair enough to show the trend of the changing air quality. Next, the

air sensing information needs to be transmitted and processed in time. It is well known that as a short-
range transmission technique, WSNs can only provide very limited coverage, which cannot meet the

communication requirements of a massive number of air quality sensors over a large sensing area.

To address the above problems, low power wide area (LPWA) technology, an emerging Machine-to-

Machine (M2M) communications technique may be used. It is specifically designed to provide ubiquitous

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

coverage, low energy consumption and costs for devices that infrequently send or receive messages with

limited data rates [10] [11] [12] [13]. In general, an LPWA network can provide suburban and rural ranges
of over 20 kilometer (km), and a typical urban range of around 5 km. Moreover, LPWA technology can also
enable devices to operate with a long battery life for over ten years. These features of LPWA technology

are particularly attractive for typical smart city applications such as urban air quality monitoring. Through
integrating LPWA technology with the air quality monitoring system, air quality monitoring nodes can
be widely scattered across every corner of the urban areas to greatly improve the spatial and temporal

resolutions of monitoring. Furthermore, the star topology and long battery life of LPWA networks can
also reduce maintenance and deployment costs of the air quality monitoring system. To the best of our

knowledge, there are very few studies on LPWA-based IoT applications in the existed literatures [14].

This paper proposes an air quality monitoring system based upon LPWA networks. Firstly, we briefly

present the architecture of the proposed system. Then, the hardware and software designs are discussed in

detail. The LPWA access point (AP) is implemented on an open-source soft-defined radio (SDR) platform.

The air quality monitoring node consists of sensors, microcontroller unit (MCU), battery and so on, which

are developed by ourselves. In order to process and analyze the sensed data, the IoT cloud with different

types of servers has been established. The users can access the air quality information either through

a website or a mobile application (APP). The proposed air monitoring system has been deployed and

operated, which demonstrates the effectiveness of the proposed system in urban environments.

II. S YSTEM A RCHITECTURE

As illustrated in Fig. 1, a three-tier hierarchical IoT architecture is adopted in the proposed air quality
monitoring system, i.e., the sensing layer, network layer and application layer.

A. Sensing layer

This is the foundation of the entire air quality monitoring system, which is mainly responsible for air
quality sensing and data reporting. As the main entities of the sensing layer, the air quality monitoring
nodes are battery powered and widely installed across a large geographical area. Thus, large-scale air

quality data can be collected by these nodes. The details of the monitoring node are provided in the next
section.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

Application layer

API API
Data Processing Storage Server
HTTP Server
Server
Web APP
Cloud
API

Network layer

GPP GPP GPP

USRP USRP USRP

LPWA AP LPWA AP LPWA AP

Sensor layer

Monitoring
Node

Fig. 1. Architecture of the LPWA-based air quality monitoring system.

B. Network layer

An LPWA network based on the IEEE 802.15.4k specification is employed to provide ubiquitous

connectivity between the monitoring nodes and the Access Point (AP). To achieve excellent coverage

performance, a particular direct sequence spread spectrum (DSSS) technique is adopted in the physical
layer (PHY), which uses a large spreading factor (SF) to achieve excellent receiver sensitivity. In our

system, the AP is developed based on open source software-defined radio (SDR) due to its flexible and

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

inexpensive rapid prototyping.

C. Application layer

The application layer is imperative for processing sensed data and providing interactive services to
users. It can be divided into two parts, i.e., IoT cloud, and client applications. Once the AP receives
the air quality data reported from the sensor nodes, it firstly stores the data in the database of the IoT

cloud. Then, the collected air quality data are aggregated, processed and analyzed in the IoT cloud.
Moreover, the IoT cloud also supports information delivery and service creation, which can be triggered
by client applications, either a computer-based web display system or smartphone-based mobile APP, via

the application programming interfaces (APIs). With the aid of client applications, users can easily obtain

the air quality information from the IoT cloud.

III. S YSTEM I MPLEMENTATION

A. Hardware Implementation

The hardware of the system mainly includes: 1) Monitoring Node; 2) Access Point (AP) Node.

Monitoring node is used to collect real-time air quality data and transmit these data to the AP through

wireless channel. The AP is responsible for receiving wireless signals and detecting the air quality data

accurately. Detailed introduction about these two components is presented as follow.

1) Monitoring Node: As shown in Fig. 2, each monitoring node consists of four functional components,

i.e., the sensor module, controller module, LPWA transmitter module and power module. The sensor
module is used to sense air quality information and transfer them to the controller module through
the Universal Synchronous/Asynchronous Receiver/Transmitter (USART) interface. Then, the controller

samples the air quality data and transmits them via the radio-front (RF) module, whose baseband signal
processing procedures are implemented according to the IEEE 802.15.4k specification [15]. Also, it is
responsible for performing all the control workflow in the node. The monitoring node is powered by a

solar battery system that supports long-term use.

a) Sensor Module: It is usually composed of several typical sensors including the particulate matter

sensor, and the temperature and humidity sensors. Compared to the ultra-red types, the laser PM2.5 air

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

LPWA
Battery Solar Tx Module
Charge
Solar Panels Controller
RF CC1125
Power Module

Control Functions Baseband Signal

Process Functions

Controller Module

Temperature
PM Sensor Humidity sensor
Sensor
Sensor Module

Fig. 2. Main functional components of the monitoring node.

particle/dust sensor is of high precision and can be easily used without calibration. Thus, the PMS5005

sensor is chosen in our monitoring node. When particles go through the detecting area in the PMS5005

sensor, light scattering is induced and the scattered light is transformed into electronic signals. Then,

the signals are amplified and processed to obtain the number and diameter of the particles, which have

certain correlation with the signal waveform [16]. SHT20 is optionally used, which has a capacitive-type

humidity sensor, a band-gap temperature sensor, and specialized analog and digital integrated circuits. It

can provide calibrated, linearized sensor signals in digital I2C format.

b) Controller Module: All control functions in the control module are implemented in the STM32F103RC

Microcontroller Unit (MCU), which is an ARM 32-bit Cortex-M3 micro-controller operating at 72 MHz
frequency, high-speed embedded memories (Flash memory up to 512 Kbytes and SRAM up to 64 Kbytes),

and an extensive range of enhanced I/Os and peripherals connected to two APB buses [17]. This MCU

offers three 12-bit ADCs, four general-purpose 16-bit timers plus two PWM timers, as well as standard and
advanced communications interfaces, i.e., up to two I2Cs, three SPIs, two I2Ss, one SDIO, five USARTs,

one USB and one CAN. These features make STM32F103RC suitable for our proposed systems.

The monitoring node is in the RUN mode after power on or reset. In order to reduce power consumption,
it usually stays in the low power mode until woken up either periodically or event-evoked, and then
enters the RUN mode. There are usually three low power consumption modes, i.e., the SLEEP, STOP and

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

STANDBY modes. In the SLEEP and STOP modes, all I/O pins keep the same state as those in the RUN

mode, and the contents in the SRAM and register are kept, which consumes power to a certain extent.
Moreover, the STANDBY mode needs the least power but loses the contents stored in the SRAM and
register, which is preferred to be used in our design.

On the other hand, it may take much time and power consumption to calculate the spreading sequences
with the SF up to 32768 for transmission in an LPWA system. After the node is powered on, the spreading

sequence is only calculated once with the given user identification and then stored in the flash memory,
whose contents can be kept in any mode. Then, after the node is woken from the STANDBY mode, the

spreading sequences can be read for use without additional computation.

To further save the energy, adaptive duty cycle adjustment is employed in the node. Since the air quality

usually does not vary too much, the node is woken to sense the air quality and transmit data every ten

minutes by default. If the difference between two successive blocks of sensing data is larger than the

given threshold, the duty cycle is lowered so as to monitor the air quality in time. Otherwise, the sleeping

duration may be increased to the default value.

c) LPWA Transmitter Module: The baseband signal processing functions of the LPWA transmitter are

first programmed in the MCU. Next, the GFSK modulated signal is transmitted through the RF chip, i.e.,

Texas Instruments (TI) CC1125 chip, which is a fully integrated single-chip radio transceiver designed

for high performance operations with very low-power and low-voltage. It has very excellent receiver

sensitivity, i.e., -129 dBm at 300 bps, -123 dBm at 1.2 kbps and -110 dBm at 50 kbps, respectively. The
ISM band of 433 MHz is used for transmission with a maximal transmit power of 15 dBm [23].

d) Power Module: The node is lithium battery powered, which can be recharged via a solar panel.
One of the problems with solar power is that the output voltage of the solar panel is often variable and
too large for charging the battery. The solar charge controller is responsible for reducing the voltage and

storing the electric energy in the battery safely. Furthermore, it can also monitor the battery level in real
time so as to protect the battery from over-charging or over-discharging. With the help of the solar charge
controller, the battery can provide a safe and stable power supply for each module of the monitoring node.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

Parallel Processing

Convolutional De-
De-Spreading
Decode Interleaver

Preamble
Demodulator
Detection

Timing USRP B210

CRC Check
Recovery

PC

Fig. 3. Functional diagram of the AP.

2) Access Point (AP) Node: As shown in Fig. 3, the AP in the network layer is implemented by using a

general purpose processer (GPP) based open source SDR platform, which consists of a common computer
and a universal software radio peripheral (USRP). An Ettus USRP B210 is chosen as the general RF unit
due to its wide frequency range from 70 MHz to 6 GHz covering all the operating frequency bands of

IEEE 802.15.4k. After down frequency conversion and digitization in USRP B210, the received signal

is transferred to the common computer with GPP via the USB 3.0 interface. Then, the baseband signal

processing functions of the receiver are realized using the real-time GNU radio-based signal processing

framework. At the receiver, non-coherent detection is first used for symbol demodulation to eliminate

the effect of the frequency and phase offsets. Next, a parallel preamble detection scheme is adopted to

handle concurrence data transmission from a massive number of monitoring nodes [10]. In each processing

thread, a fast Fourier transform (FFT)-based algorithm is employed to implement the correlation process

for preamble detection, and a dynamic timing adjustment algorithm is used to track the optimal sampling

time [24].

B. Software Implementation

As shown in Fig. 1, a large number of software programs are required for not only in the IoT cloud but
also the client so as that users can make the full use of the services provided by the processed system.

1) Server: There are mainly three servers in the IoT cloud with different functions, i.e.,

• Data Processing Server

The sensed air quality data arrive at the IoT cloud continuously. Raw data cannot be directly utilized

due to the presence of possible transmission errors and machine failures. Specific pre-processing

procedures are needed to detect and filter out “dirty” data so as to ensure data integrity and reliability.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

Next, the filtered data are analyzed to reveal the air quality trend by the data processing server. For

example, when a user wants to know the air quality information of a particular geographical location
during the past one week or even one month, the processed data may be sent instead of the raw data.
By this way, a small amount of data is needed for transmission to reduce latency.

• Storage Server
Not only the filtered raw data but also the analytical results are stored in the database of the storage
server. Moreover, there is a trigger function in the storage server to notify the user when new analytical

results become available. As soon as new sensing data arrive at the cloud, new analytical results will
be generated by the data processing server and stored in the database. Meanwhile, it is also sent

to the active user with the WebSocket protocol. WebSocket is applied in order to provide efficient

and reliable communication between the server and clients. Without repeated HTTP headers, the

transmission delay of air quality data can be reduced to a certain extent [18]. In this way, the server

is able to update the air quality information in real time while avoiding too many simultaneous

requests.

• HTTP Server

The HTTP protocol is used to provide services to users, and thus an HTTP server is developed using

Servlet and JSP (JavaServer Pages) [19]. It is deployed in a web container, i.e., Tomcat, which is

responsible for managing the lifecycle of servlets. The HTTP server interacts with clients through

a two-way Request-Response mode, namely POST and GET. Also, it provides APIs for clients to

request the air quality information by using a web browser or a mobile APP.

2) Display Applications: As aforementioned, either a website or a mobile APP can be used to display

the air quality information to end users. Our homepage is developed using HTML, JavaScript and CSS,

and an Android APP is also provided. Fig. 4(a) and Fig. 4(b) illustrate the GUI interfaces of the website
and the mobile APP, respectively.

Through the display applications, the air quality information can be shown in real-time, e.g.,

• Current Air Quality Indicator (AQI);

• AQI trend of the present day;

• AQI trend in last week/month.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

10

(a) Webpage.

(b) Mobile APP.

Fig. 4. Examples of GUI interface in the air quality monitoring system.

IV. E XPERIMENTAL R ESULTS AND A NALYSIS

A. Experimental Configurations

Our experiments are carried out in a typical urban environment. The wipe antenna of the AP is installed
on the roof of a 15-floor building at the center of the university campus. The received RF signal is fed

to USRP B210 located in the room through a 10-meter cable. The main system parameters in field tests

are listed in Table I. Our LPWA-based air quality monitoring system is deployed using a star topology
as illustrated in Fig. 5(a), which consists of five monitoring nodes distributed over an area of around a

radius of 3 km from the center of university campus. The photo of a monitoring node is shown in Fig.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

11

TABLE I
M AIN PARAMETERS OF THE LPWA NETWORK .

Parameters Value
PHY Mode DS-SS
Modulation scheme GFSK
GFSK deviation 100 kHz
Carrier frequency 433 MHz
Symbol rate 200 ksym/s
Preamble 4 bytes
PSDU 16 bytes
Spreading factor (SF) 32768
Transmit power 5 dB, 15 dB
Number of Sensing nodes 5

5(b). Both the AP and monitoring node operate at the frequency band of 433 MHz with a symbol rate 200

Kbps. In each packet sent by the monitoring node, the physical protocol data unit (PPDU) that consists of

a 4-octet preamble and a 16-octet physical service data unit (PSDU) is spreaded by the Gold code with

an SF of 32,768. Each monitoring node is powered by a solar panel and reports the collected air quality

data to the LPWA AP every 10 minutes.

As a metric to indicate air quality, the Air Quality Index (AQI) α is calculated through measuring six

major air pollutants, i.e., PM2.5 , PM10 , carbon monoxide (CO), sulfur dioxide (SO2 ), nitrogen dioxide

(NO2 ), and ozone (O3 ) [20]. The higher the AQI, the greater the level of air pollution and as well as the

health concern. The level of each pollutant, i.e., Individual Air Quality Index (IAQI), is computed as

Ahigh − Alow
ip = (βp − Blow ) + Alow , (1)
Bhigh − Blow

where ip is the IAQI of the pth pollutant, βp is the concentration of the pth pollutant, Bhigh is the nearest
concentration point larger than βp , Blow is the nearest concentration point smaller than βp , Alow is the
IAQI point corresponding to Blow , and Ahigh is the IAQI point corresponding to Bhigh .

The critical points for six air pollutants are given by China’s Ministry of Environmental Protection in

[20]. For example, the critical points for PM2.5 and PM10 are listed in Table II, which are the only two

metrics concerned in our air quality monitoring system. Then, the AQI can be given as the highest value
of all the measured IAQIs, i.e.,

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

12

Node 3
[5 dBm]
Node 2
3 km 2 km 1 km [5 dBm]
Node 1
[5 dBm]

Node 5 Node 4
[15 dBm] [15 dBm]

AP Node

(a) Location of the monitoring nodes in the field trial.

Solar Panel

Monitoring
Node

(b) Photo of the monitoring node.

Fig. 5. Deployment of the LPWA-based air quality monitoring system in the field trial.

α = max{i1 , i2 , ..., in }. (2)

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

13

TABLE II
C ONCENTRATION POINTS FOR PM2.5 AND PM10 .

IAQI PM2.5 (µm/m3 ) PM10 (µm/m3 )

0 0 0
50 35 50
100 75 150
150 115 250
200 150 350
300 250 420
400 350 500
500 500 600

B. Results and Analysis

To evaluate the performance of the LPWA network, we utilize the carrier-to-interference (C/I) value of

received packets as the performance metric, which is defined as the ratio of the maximum of normalized

correlation values calculated in the preamble detector to the received signal power. When a node is

deployed, its transmit power is configured according to its distance to the AP, e.g., 5 dBm for the distance

within 0.5 km, and 15 dBm for the distance between 0.5 km to 3 km. Thus, without excessive power

consumption, the C/I values of all the nodes are almost larger than -30 dB, which is the C/I threshold

to ensure that the received packets are decoded successfully [10]. Meanwhile, the impact of the near-far

effect can be mitigated too. In order to demonstrate the effectiveness of our system design, Fig. 6 plots

the average C/I value of the received packets from different monitoring nodes, where the average C/I

values of all the nodes are higher than the C/I threshold. As can be seen from Fig. 6, for N odes 1, 3,

4, and 5, the average C/I value gradually decreases with the increase of the node’s distance to the AP. It
is because a longer distance generally results in a larger path loss, which has an immediate effect on the

computation of the correlation values in the preamble detector of the AP. Although N ode 2 and N ode 3

are deployed in nearly the same distance to the AP with the same transmit power, N ode 2 has a much
lower average C/I value than N ode 3 by 20 dB. This is because a building is located along the direct

path between N ode 2 and the AP, so that N ode 2 suffers from more severe channel fading.

Air quality data are collected through the LPWA system and used for further analysis. Only the measured
data from April 1-14, 2016 are chosen for the sake of illustration. Air quality data including PM2.5 , PM10

and AQI is provided in Fig. 7. In order to verify the reliability of this air quality monitoring system,
another set of data is obtained from a well-known PM2.5 historical database, i.e., the Young-0.com [22].

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

14

5 Node 1
Node 3 C/I

-5

C/I (dB)
-10 Node 4

-15
Node 2
-20
Node 5
-25

Threshold= -30 dB
-30

0.0 0.5 1.0 1.5 2.0

Distance (km)

Fig. 6. Transmit power and C/I value of each node.

300
PM2.5
PM10
250 AQI

200
Air Quality (ug/m )
3

150

100

50

0
Apr. 1st 3 5 7 9 11 13 15

Fig. 7. Air quality data including PM2.5 , PM10 and AQI.

The two datasets are compared in Fig. 8. It is evident that the trends of the two lines well coincide,
and the discrepancy may be due to the different measurement locations. Therefore, the air quality data

obtained by the proposed system can satisfy both the accuracy and reliability requirements.

PM2.5 concentration is highly related to some meteorological features such as the temperature, humidity

and wind speed, whose data can be obtained from a meteorological data sharing website [21]. As an
example, Fig. 9 plots the relationship between PM2.5 and the wind speed. From April 3-7, 2016, the
PM2.5 concentration increased gradually with the wind speeds slower than 4 m/s, and reached the peak

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

15

350
Measured Data
Reference Data
300

250

PM2.5 (ug/m )
200
3

150

100

50

0
Apr. 1st 3 5 7 9 11 13 15

Fig. 8. Comparison between the measured and reference data on PM2.5 concentration.

240 8
PM2.5
Wind Speed
7
200

160

Wind Speed (m/s)

5
PM2.5 (ug/m )
3

120 4

3
80

40
1

0 0
Apr. 1st 3 5 7 9 11 13 15

Fig. 9. Relationship between PM2.5 concentration and the wind speed.

on the noon of April 7, 2016. Then, it dropped rapidly below 20 µg/m3 and kept low during April 8-11,
2016, which was due to the dispersal of air pollutants by strong winds. Thus, it can be concluded that air

quality depends on the variation of the wind speed somehow.

V. C ONCLUSION

We have implemented an air quality monitoring system by using the advanced IoT techniques in this

paper. With the aid of the LPWA network, the air sensing data over a large coverage area is collected

and transmitted to the IoT cloud in time. The portal monitoring nodes are developed for easy deployment

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

16

and can work all day with a battery or a solar panel. All the functions of the AP are implemented on

a GPP-based SDR platform. The sensed data are stored in the database and analyzed in the IoT cloud.
A plenty of experiments have been carried out in the urban environments to validate the reliability of
the proposed system. Some interesting facts have been revealed when comparing the air quality trend

and other similar data. It is believed that long-term and large-scale air monitoring can greatly help us
understand air pollution and find a way to solve the problem of air pollution at least partially.

VI. ACKNOWLEDGMENT

This work was supported by the National High-Tech R&D Program (863 Program 2015AA01A705),

the National Key Technology R&D Program of China under the grant 2015ZX03002009-004, the China

Natural Science Funding under the grant 61271183, and the Fundamental Research Funds for the Central

Universities under grant 2014ZD03-02.

R EFERENCES
[1] E. Fotopoulou, A. Zafeiropoulos, D. Papaspyros, P. Hasapis, G. Tsiolis, T. Bouras, S. Mouzakitis, and N. Zanetti, “Linked data analytics
in interdisciplinary studies: the health impact of air pollution in urban areas,” IEEE Access, vol. 4, pp. 149-164, Mar. 2016.
[2] K. B. Shaban, A. Kadri, and E. Rezk, “Urban air pollution monitoring system with forecasting models,” IEEE Sensors Journal, vol.
16, no. 8, pp. 2598-2606, Apr. 2016.
[3] W. Li, K. Lo, and M. Terrence, “A survey of wireless sensor network based air pollution monitoring systems,” Sensors, vol. 15, no.
12, pp. 31392-31427, Dec. 2015.
[4] C. Peng, K. Qian, and C. Wang, “Design and application of a VOC-monitoring system based on a ZigBee wireless sensor network,”
IEEE Sensors Journal, vol. 15, no. 4, pp. 2255-2268, Apr. 2015.
[5] S. He, J. Chen, X. Li, X. Shen, and Y. Sun, “Mobility and intruder prior information improving the barrier coverage of sparse sensor
networks,” IEEE Transactions on Mobile Computing, vol. 13, no. 6, pp. 1268-1282, Jun. 2014.
[6] J. Chen, W. Xu, S. He, Y. Sun, P. Thulasiraman, and X. Shen, “Utility-based asynchronous flow control algorithm for wireless sensor
networks,” IEEE Journal on Selected Areas in Communications, vol. 28, no. 7, pp. 1116-1126, Aug. 2010.
[7] L. Lei, Y. Kuang, X. Shen, K. Yang, J. Qiao, and Z. Zhong, “Optimal reliability in energy harvesting industrial wireless sensor
networks,” IEEE Transactions on Wireless Communications, doi: 10.1109/TWC.2016.2558146.
[8] Y. Ma, S. Yang, Z. Huang, Y. Hou, L. Cui, and D. Yang, “Hierarchical air quality monitoring system design,” in Proc. IEEE 2014
International Symposium on Integrated Circuits (ISIC), Singapore, Dec. 2014, pp. 284-287.
[9] Y. Yang and L. Li, “A smart sensor system for air quality monitoring and massive data collection,” in Proc. IEEE 2015 International
Conference on Information and Communication Technology Convergence (ICTC), Jeju, South Korea, Oct. 2015, pp. 147-152.
[10] X. Xiong, K. Zheng, R. Xu, W. Xiang, and P. Chatzimisios, “Low power wide area machine-to-machine networks: key techniques and
prototype,” IEEE Commun. Mag., vol. 53, no. 9, pp. 64-71, Sep. 2015.
[11] Sigfox, Sigfox - One network a billion dreams, Whitepaper.
[12] 3GPP RAN Tdoc RP-151621, “New WI proposal: narrow band IOT,” approved in RAN plenary #69 meeting, Sep. 2015.
[13] L. Lei, Y. Kuang, N. Cheng, X. Shen, Z. Zhong, and C. Lin, “Delay-optimal dynamic mode selection and resource allocation in device-
to-device communicationspart II: practical algorithm,” IEEE Transactions on Vehicular Technology, vol. 65, no. 5, pp. 3491-3505, May
2016.
[14] S. Andreev et al., “Understanding the IoT connectivity landscape: a contemporary M2M radio technology roadmap,” IEEE Commun.
Mag., vol. 53, no. 9, pp. 32-40, Sep. 2015.
[15] “IEEE Standard for Local and metropolitan area networks Part 15.4: Low-Rate Wireless Personal Area Networks (LR-WPANs)
Amendment 5: Physical Layer Specifications for Low Energy, Critical Infrastructure Monitoring Networks,” IEEE Std 802.15.4k-2013,
Aug. 2013.
[16] X. Li, E. Iervolino, F. Santagata, J. Wei, C. A. Yuan, P. M. Sarro, and G. Q. Zhang, “Miniaturized particulate matter sensor for portable
air quality monitoring devices,” in Proc. IEEE SENSORS 2014 Proceedings, Valencia, Nov. 2014, pp. 2151-2154.
[17] Alldatasheet.com, “STM32F103RC Datasheet(PDF) - STMicroelectronics,” 2016. [Online]. Available:
http://www.alldatasheet.com/datasheet-pdf/pdf/231959/STMICROELECTRONICS/STM32F103RC.html. [Accessed: 05- May- 2016].

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2582153, IEEE Access

17

[18] V. Pimentel and B. G. Nickerson, “Communicating and displaying real-time data with WebSocket,” IEEE Internet Computing, vol. 16,
no. 4, pp. 45-53, Aug. 2012.
[19] Tomcat.apache.org, “Servlet and JavaServer Pages API documentation,” 2016. [Online]. Available: https://tomcat.apache.org/tomcat-
4.1-doc/servletapi/. [Accessed: 30- Apr- 2016].
[20] People’s Republic of China Ministry of Environmental Protection, “People’s Republic of China Ministry of Environmental Protection
Standard: technical regulation on ambient air quality index,” pp. 2, Feb. 2012.
[21] Rp5.ru, “Weather for 243 countries of the world,” 2016. [Online]. Available: http://rp5.ru/Weather in the world. [Accessed: 30- Apr-
2016].
[22] Young-0.com, “China PM2.5 report,” 2016. [Online]. Available: http://www.young-0.com/airquality. [Accessed: 30- Apr- 2016].
[23] Ti.com, “CC1125 online datasheet,” 2016. [Online]. Available at: http://www.ti.com/product/CC1125/datasheet. [Accessed 15- Jun-
2016].
[24] X. Xiong, T. Wu, H. Long, and K. Zheng, “Implementation and performance evaluation of LECIM for 5G M2M applications with
SDR,” in Proc. IEEE 2014 Globecom Workshops (GC Wkshps), Austin, TX, Dec. 2014, pp. 612-617.

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.