INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 423
www.rsisinternational.org
Advancements and Frameworks in IoT-Based Air Pollution
Monitoring Systems
Yash Gaonkar
1
;
Prof. Siddhant Jalmi
2
1
Student, Dept. of ECE
Agnel Institute of Technology and Design, Goa,India
2
Assistant Professor, Dept. of ECE
Agnel Institute of Technology and Design,
Goa, India
DOI: https://doi.org/10.51583/IJLTEMAS.2026.15020000038
Received: 16 February 2025; Accepted: 23 February 2026; Published: 05 March 2026
ABSTRACT
The Internet of Things (IoT) has emerged as a key technology for scalable, real-time environmental monitoring,
addressing the growing global challenge of air pollution. This review synthesizes findings from more than twenty
major deployments across urban, industrial and smart-city environments, including implementations from
Hindustan Institute of Technology [1], Sathyabama Institute [2], Taiwan’s integrated governance framework [3],
the Bulgarian Academy of Sciences [4] and low-cost deployments in developing regions. Comparative analysis
spans sensing hardware, communication technologies (WiFi, LTE/4G, LoRa), cloud architectures, machine
learning integration and field performance. Results indicate significant improvements in spatial coverage, real-
time responsiveness and data-driven governance, while highlighting persistent challenges in sensor calibration,
cybersecurity and long-term reliability. Emerging advances in edge intelligence, spectral sensing, federated
learning and hybrid communication architectures are examined as pathways toward next-generation
environmental monitoring. This consolidated review provides a structured framework for scalable, resilient and
policy-integrated IoT air-quality monitoring systems.
Keywords: IoT, air pollution monitoring, environmental sensors, real-time data, machine learning, smart
cities, sustainability, cloud computing
INTRODUCTION
Air pollution is a major environmental and public health threat, contributing to approximately 4.2 million
premature deaths annually worldwide. Rapid urbanization, industrial growth and increased vehicular emissions
have intensified environmental degradation, particularly in densely populated metropolitan areas. Traditional
monitoring systems rely on centralized stations, offering limited spatial coverage, delayed reporting and high
operational costs, restricting accessibility and real-time response [19][20].
IoT-based monitoring enables distributed sensing, low-power wireless communication, cloud analytics and
machine learning, providing continuous, real-time and spatially dense environmental observation. Compared
with legacy systems, IoT architectures deliver immediate data availability, scalable deployment and improved
decision-making across municipal, regional and national governance levels [19][20].
This review synthesizes empirical evidence from multiple deployments, including Hindustan Institute of
Technology and Science [1], Sathyabama Institute [2], Taoyuan smart city infrastructure [3], Bulgarian reliability
studies [4] and community-driven low-cost systems [6]. The objective is to identify technological evolution,
comparative performance, documented outcomes, persistent challenges and future research directions.
SYSTEMATIC LITERATURE REVIEW METHODOLOGY
This review follows a systematic approach aligned with PRISMA guidelines to ensure comprehensive and
reproducible selection of studies. Databases searched include IEEE Xplore, Scopus, Google Scholar, and
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 424
www.rsisinternational.org
SpringerLink (January 2018–October 2025). Keywords: 'IoT air pollution monitoring', 'IoT environmental
sensors', 'smart city air quality IoT', 'LoRa/WiFi air pollution'. Inclusion criteria: peer-reviewed English papers
on empirical IoT deployments with hardware/comms/ML details, > 6 months field testing. Exclusion:
simulations only, non-IoT sensors.
Hardware Architectures and Sensor Technologies
Microcontroller
Platforms
IoT air pollution monitoring systems employ diverse microcontroller architectures optimized for cost, power and
processing capability.
Arduino / ATmega Series: The ATmega328P platform, used in Hindustan and Sathyabama deployments,
provides 32 kB flash, 2 kB SRAM, 1 kB EEPROM, 23 GPIO pins, 10-bit ADC and serial communication
interfaces (UART, SPI, I²C). Low-power operation supports battery-based deployments and its cost (≈$2–5)
enables scalable use in developing regions [1][2].
NodeMCU ESP8266: The ESP8266 integrates WiFi (802.11 b/g/n), up to 17 GPIOs, 10-bit ADC and clock
speeds up to 160 MHz. Eliminating external communication modules, it enables compact real-time IoT nodes
widely deployed in India, Senegal and Coimbatore [6][7].
Raspberry Pi and Advanced Platforms: Industrial and smart-city deployments employ Raspberry Pi-class single-
board computers with multi-core processors, 1–8 GB RAM and full Linux operating systems, enabling edge
analytics, local preprocessing and complex system integration [3][4].
Gas Sensor Technologies
Sensor selection critically determines system capabilities, accuracy and cost- effectiveness.
MQ-135 Multi-Gas Sensor: Deployed across Hindustan Institute, Sathyabama and international
implementations the MQ-135 detects CO
2
, NO
2
, NH
3
, benzene, smoke and other pollutants. Operating on tin
dioxide (SnO
2
) semiconductor principles with a heated ceramic sensing element, it responds to gas exposure
through changes in electrical conductivity.[1][2]
Specifications include:
Detection range: 10 ppm to 10,000 ppm
Response time: $<$10 s
Operating temperature: −10 °C to 50 °C
Sensitivity adjustable via potentiometer
Output: Analog voltage (0–5 V) proportional to gas concentration
Cost: $3–5 per unit
MQ-7 Carbon Monoxide Sensor: Designed for CO detection with a typical range of 20–2000 ppm; used in
traffic pollution assessment.
MQ-5 and MQ-6 Gas Sensors:
MQ-5 detects LPG, methane and natural gas (200–10,000 ppm).
MQ-6 targets LPG, isobutane and propane; common in industrial and vehicle monitoring.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 425
www.rsisinternational.org
Environmental Integration Sensors: DHT11/DHT22 humidity and temperature sensors (approximately ±2
◦
C
for temperature, ±5% for humidity) provide critical context variables affecting sensor readings and air quality
interpretation.
Sensor Calibration and Accuracy
A persistent challenge for air pollution monitoring across IoT implementations lies in ensuring long-term sensor
accuracy and reliability. Field reports from Hindustan Institute and Sathyabama consistently documented
deviations as high as ±10% [1][2] for commodity gas sensors in varied environmental con- ditions. This drift
arises from changes in humidity, temperature and physical fouling compounding over months of operation and
eroding confidence in longi- tudinal datasets. More recent research from the Bulgarian Academy of Sciences
employed Bayesian Deep Belief Neural Networks (BDBN) for auto-calibration, achieving reliability scores
exceeding 95% [4]. Best practices for reliability, as compiled from international studies, include routine cross-
validation against reference stations, developing firmware algorithms for temperature/humidity compensation,
scheduled periodic recalibration and deployment in robust enclo- sures.[4][30]
Data Transmission and Communication Pro- Tocols
WiFi-Based Urban Deployments
WiFi remains the transmission backbone for most urban IoT air monitoring networks. Its popularity is due to
widespread municipal infrastructure, high bandwidth (enabling transmission of rich, high-frequency datasets)
and low la- tency with many systems reporting latencies comfortably below 100 ms. Secu- rity protocols like
WPA2 ensure reasonable protection against data interception and unauthorized access. Moreover, compatibility
with existing city networks makes integration cost-effective.
WiFi-aware deployments are not without problems. Urban environments present numerous “dead zones” owing
to building density and interference and data reliability suffers in areas more than 50–100 meters from public
routers. Power draw during active transmission—often around 80 mW is nontrivial, im- pacting battery longevity
unless nodes are plugged into mains or augmented with solar charging. Lastly, system throughput and stability
can drop sharply during periods of network congestion, restricting reliable monitoring in mega- cities or areas
of rapid population growth.[8]
LTE/4G Cellular Networks
Industrial, mobile, and vehicular networks often pivot to LTE/4G protocols for their enhanced geographic
coverage and seamless mobility. These systems can support data collection from devices mounted on
motorcycles, cars and indus- trial plants spread across broad areas. Integration with regulatory databases is an
emerging trend, promising near real-time regulatory enforcement and dy- namic threshold management.
Despite these strengths, cellular deployments introduce higher per-node op- erational costs—commonly ranging
from $5 to $15 per device monthly—and depend heavily on regional cellular provider availability. They also
tend to ex- perience fluctuating latency (from 50 to 200 ms), which may lag behind WiFi for highly granular,
time-sensitive measurements.
LoRaWAN Long-Range Protocol
Deployments in rural and industrial contexts increasingly favor LoRaWAN for its ultra-low power consumption
(often below 10 mW during transmission) and impressive spatial reach (up to 10 km in open terrain, 1–3 km in
urban settings). LoRa enables periodic, low-bandwidth readings at minimal cost, which is ideal for distributed
sensor networks in resource-constrained environments.
However, LoRa’s bandwidth limitations mean payloads are small and trans- mission frequency is low, reducing
temporal resolution and making it less suit- able for rapid-response urban air quality tracking. Nevertheless,
among large- area deployments with cost constraints, LoRa delivers robust performance.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 426
www.rsisinternational.org
Cloud Integration and Data Transmission Protocols
Modern IoT platforms coordinate device-to-cloud data flows using protocols like MQTT and secured
HTTP/HTTPS. The typical operational pipeline first preprocesses and calibrates raw sensor data on-device (Edge
Preprocessing), then transmits via MQTT to brokers hosted by providers such as AWS, Google Cloud, or
ThingSpeak (publish–subscribe mechanism).
Data is stored in cloud time-series databases with all relevant metadata; finally, web dashboards, mobile apps,
or public APIs visualize the measurements for stakeholders. Specific platforms such as Blynk streamline
integration and support threshold-based alerting, further enhancing utility for researchers and citizens.[9][10]
Data Analytics and Machine Learning Integration
Urban Networks:Hindustan Institute and Sathyabama Projects
The Hindustan Institute’s Chennai-based deployment (2021–2022) demon- strated the benefits of close-knit
sensor networks, combining CO, CO
2
, NO
2
, temperature and humidity monitoring via MQ-7 and MQ-5 sensors
linked to Arduino Uno microcontrollers and ESP8266 modules. Real-time monitoring was achieved, with all
sensor readings reliably transmitted to Google Cloud’s Firebase and visualized on mobile apps. Following local
awareness campaigns and system data utilization, PM2.5 concentrations dropped by 15–22% in moni- tored
neighborhoods. Uptime was exceptional (>98%), and the sensor readings remained within ±8–12% deviation
against reference stations, reinforcing the value of regular recalibration and robust system architecture.[1]
Sathyabama Institute’s multi-node Chennai network similarly scaled from 12 to 15 sensor nodes, achieving
comprehensive multi-pollutant tracking (CO, CO
2
, NO
2
, SO
2
) across residential, commercial and industrial
zones. Data ag- gregation leveraged WiFi and cloud storage, supporting a data delivery success rate above 90%
and consistent sensor deviation below 10%. The system’s cost- effectiveness (approx. $150–$200 per node)
enabled city-wide expansion, a point validated through project scalability pilots.[2]
Smart City Integration: Taoyuan City, Taiwan
System Architecture: The Taoyuan smart-city IoT air-quality system consists of over 200 distributed multi-gas
sensor nodes deployed across industrial, residential, and traffic zones. A hybrid LTE–WiFi communication
backbone provides redundancy and resilient data transfer. Each node performs local preprocessing before
transmitting data to a centralized municipal platform, where cloud analytics detect threshold exceedances, trends,
and pollution hotspots. Automated workflows notify regulators and stakeholders in near real time, enabling
integration with public dashboards, regulatory databases, and environmental health systems [3].
Documented Outcomes: The Taoyuan deployment has been extensively documented in both technical and high-
impact environmental science journals. Over a five-year period, annual average PM2.5 concentrations decreased
dramatically from 42 to 28 µg/m
3
, representing about a 33% reduction directly attributed to the data-driven
interventions supported by the IoT network. Si-multaneously, citywide NO levels fell by approximately 15% as
identified in time-series analyses spanning before-and-after implementation periods. Public satisfaction surveys,
tracked annually, revealed strong improvements, with approval of air quality management strategies rising from
42% to 71% between 2018 and 2023.
Equally important, policy response times to acute pollution events shortened drastically: whereas pre-IoT
regulatory action often involved multi-month lags, current workflows now support city-level interventions within
24–48 hours—sometimes same-day for high-priority situations.[3]
Integration with Urban Planning: Real-time sensor data supports adaptive traffic rerouting during pollution
spikes and automated industrial compliance monitoring. Emission exceedances trigger regulatory enforcement
and fines. Health integration correlates pollution data with respiratory admissions for early-warning capability.
Spatial analytics guide urban greening and pollution mitigation planning, demonstrating direct linkage between
IoT sensing and city-level environmental governance [3]
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 427
www.rsisinternational.org
Industrial Monitoring: Dr NGP Institute and Manu- facturing Zones
The industrial air pollution monitoring system developed at Dr NGP Institute (Coimbatore), in line with
contemporary best practices for industrial IoT, uti- lizes an array of SO
2
, CO, NO
2
and particulate sensors
positioned at key facility outlets to ensure continuous, real-time emissions surveillance. Each sensor unit is
connected via a LoRa or LTE network to centralized industrial monitoring centers, offering both high spatial
coverage and robust resilience to communica- tion disruptions—essential in scattered, interference-prone factory
environments. The data is not simply logged; embedded machine learning algorithms analyze emissions patterns
and predict compliance risks and the entire system is deeply integrated with local and municipal regulatory
databases.[7]
The system’s operational outcomes have been impressive. Toxic emissions are detected in real time, allowing
for rapid notification and preventative action: for example, policy response metrics show an average of just 15–
30 minutes from exceedance to action, outpacing traditional methods by an order of magnitude. Within the initial
six-month deployment window, 45 major facilities received automated corrective notices based on detected out-
of-bounds emissions events. Most notably, the overall compliance rate among participating industries rose
dramatically: baseline rates (62%) jumped to 89% post-implementation, at- tributed largely to the transparent,
data-driven feedback loops, frequent report- ing and automated alerting enabled by the IoT system. Health
outcomes were also recorded, with a 32% reduction in local respiratory complaints in com- munities adjacent to
monitored industrial zones evidence that such emission mitigation can have rapid, meaningful public health
benefits.[7]
Low-Cost Community Solutions: Africa and South Asia
In parallel to industrial systems, much recent research focuses on democratiz- ing air quality data through low-
cost, community-centric monitoring initiatives. Typical system designs leverage cheap, reliable components
such as the ESP8266 microcontroller, MQ-135 and DHT11 sensors and locally-sourced LCDs and pro- tective
enclosures, for an estimated total hardware outlay of just $40–$60 per unit. Open-source firmware, commonly
based on the Arduino IDE, guaran- tees adaptability and local maintainability, while ThingSpeak (or similar)
dash- boards provide real-time visualization and threshold-based alert notifications on both computers and
mobile devices—crucial for timely public health response in underserved regions.[6]
The results are strong on several dimensions. Community-managed nodes have demonstrated 18–24 months of
stable operation, with data collection rates consistently between 85–92% despite occasional power or signal
outages. Unlike top-down deployments, public access is a central design goal: open dashboards typically see
200–400 active users per monitoring zone and local stakeholders are routinely consulted for software upgrades
and maintenance scheduling. Be- havioral impacts have also been documented—citizen awareness of air
pollution hazards has risen noticeably, with observable changes in daily activity patterns (e.g., avoiding outdoor
exposure during local spikes), suggesting data not only informs but measurably protects public health. Most
importantly, the local capacity for low-cost repair and replacement ensures sustainability and fosters tech
stewardship within the community, closing the loop between innovation, adoption and lived experience.[6]
Comparative Analysis of System Architectures
Dimension
Urban WiFi
Smart City LTE
Industrial LoRa
Primary Platforms
Arduino, NodeMCU
Raspberry Pi, Industrial
PLC
LoRa gateway,
custom boards
Cost per Node
$150–250
$500–900
$250–500
Deployment Den- sity
> 20 per km
2
0.5–1 per km
2
2–10 per 50 km
2
Data Latency
< 100 ms
50–200
ms
2–10
minutes
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 428
www.rsisinternational.org
Communication Range
50–100
m
(indoor)
City-wide (>100 km)
1–10 km (rural)
Power Consump- tion
0.5–1 W
> 5 W
< 0.2 W
Scalability
High(WiF, infrastructure)
Very High (cellular)
High (sparse/industrial)
Data Accuracy
±5%
<1%
±10%
Government Inte- gration
Limited
Extensive municipal
Industrial compliance
ML Implementa- tion
Decision trees, basic
Deep learning
(SDNN, LSTM)
Threshold, ensemble
Health Impact
Moderate
High (PM2.5 reduction
20–30%)
High (incidents pre-
vented)
Table 1: Comparative summary of IoT architectures from PRISMA-selected studies (n=30), highlighting
platforms, performance, and outcomes."
Machine Learning Performance and Reliability
BDBN Algorithm Performance in Reliability Analysis
A significant innovation in reliability assessment for IoT air pollution monitoring has been introduced by the
Bulgarian Academy of Sciences, who developed and validated a Bat-based Deep Belief Neural Network (BDBN)
framework tailored for risk factor prediction in air quality data. This approach was tested rigorously under a
range of challenging network and sensor error scenarios, representing real-world environmental conditions.
To test robustness, the evaluation involved deliberately injecting noise in the transport layer (with a simulated
packet loss rate between 5% and 15%) and introducing occasional outlier sensor readings to emulate field
anomalies. The BDBN model demonstrated remarkable resilience: when classifying the air quality state (e.g.,
“safe”, “alert”, “hazard”), it achieved a 95.3% accuracy rate, vastly outperforming traditional regression models
under these disruptive conditions. Quantitatively, the mean absolute error (MAE) for carbon monoxide
concentration predictions was just 3.2 ppm for BDBN, compared to 8.7 ppm for simple regression. Similarly,
the root mean square error (RMSE) for BDBN was 5.1 ppm, as opposed to 12.4 ppm for regression—clear
evidence of enhanced reliability and predictive stability.[4]
Correlational performance metrics further highlighted the improvement: the Pearson correlation coefficient was
r = 0.94 (indicating a very strong linear relationship between predicted and reference values) and the coefficient
of de- termination reached R² = 0.88, showing that 88% of sensor reading variance was accurately captured by
the model. The system’s overall error rate, notably the rate of air quality “state” misclassification, was reduced
to just 4.7%, an order-of-magnitude better than rule-based models.[4]
The broader literature on BDBN and related deep neural network methods for IoT sensor reliability corroborates
these findings, noting that embedded optimization and adaptive learning approaches yield dramatic gains in
resilience, self-calibrating performance and cost-effectiveness for real-world environmental monitoring
networks.
Deep Learning for Predictive Analytics
The adoption of deep learning and ensemble machine learning techniques for real- time air pollution prediction
is a rapidly maturing trend in IoT-based air quality research. Hemanth Karnati’s work, leveraging the
ThingSpeak IoT cloud and sensor data streams, provides a representative case study of these approaches in
practice. In his experiments, sensor networks comprising low-cost MQ135 and MQ3 modules captured on-the-
fly air composition across urban settings. The resulting data, stored and managed on the ThingSpeak platform,
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 429
www.rsisinternational.org
was subjected to machine learning analysis using both classical models (like Random Forest) and neural network
variants such as LSTM and GRU, which are well-suited to time-series forecasting.[5]
Model evaluations demonstrated that short-term (24–72 hour) pollution level forecasting could be achieved with
an accuracy of 82–88%, based on Mean Ab- solute Error (MAE) and Root Mean Square Error (RMSE) metrics
for AQI prediction. This performance was competitive with or superior to traditional regression and older
threshold-based approaches, particularly under complex and noisy urban data conditions. Critical for public
health, these models also excelled in anomaly detection tasks: for pollution spike events or sensor drift outliers,
the model achieved 91% sensitivity in detecting out-of-baseline events, allowing for early warnings and
automated responses (such as user alerts or actuator deployment).[5]
Importantly, the implementation design ensured that false-positive rates— the likelihood of issuing unnecessary
alerts—remained in the moderate range of 8–12%, which is considered a reasonable trade-off in operational
smart city deployments that prioritize safety and rapid notification. End-to-end latency, from sensor
measurement to cloud data processing and mobile notification, was mea- sured at roughly 3–4 seconds in
practical trials, affirming the viability of deep learning IoT platforms for near-real-time public health
interventions. These findings are further corroborated by literature synthesizing multi-sensor deploy- ments with
deep learning pipelines on open-source cloud platforms, showing that the convergence of IoT hardware, cloud
and advanced machine learning is a robust pathway for scalable, actionable urban air quality forecasting and
management[5]
Comparison with Traditional Methods
The emergence of machine learning (ML) for air quality analysis marks a sub- stantial leap beyond threshold-
based alerting systems that have long dominated field deployments. Conventional threshold approaches typically
trigger warnings when pollutant concentrations exceed fixed regulatory values, without accounting for sensor
drift, local variation, or complex pollutant interactions. This results in frequent false alarms or missed early
warnings, limiting utility for public health protection and policy response.
In contrast, ML-based forecasting and classification systems—whether based on decision trees, random forests,
or deep learning—consistently demonstrate a significant advantage in predictive performance. Comparative
studies in both academic and applied settings report that ML systems improve forecast ac- curacy by 15–25%
relative to naive threshold rules when measured by mean absolute error and out-of-sample prediction.
This boost derives from ML’s ability to capture nonlinear dependencies across variables (such as the combined
influence of meteorological parameters, emissions and seasonality) that thresh- old logic cannot model. Machine
learning approaches are also more adaptable to new environments and evolving emission patterns, a key strength
for urban deployments and smart city applications.[9]
Furthermore, advanced models - especially those employing ensemble meth- ods, deep learning architectures
like BDBN or LSTM, or hybrid approaches— demonstrate substantial reductions in the rate of false positives.
Across multiple peer-reviewed field studies and benchmark datasets, false alarm rates have been reduced by 20–
35% versus classic threshold systems. The inclusion of ML-driven anomaly detection or dynamic filtering also
allows for early identification of sen- sor faults, data drift, or extreme pollution events, a critical benefit in
practical deployments.[10]
Perhaps most importantly for operational use, real-time ML processing— whether on the sensor edge or in the
cloud—enables actionable advance warn- ing: practical deployments report early alerts 30–60 minutes before
pollution episodes reach regulatory limits, a lead time sufficient to activate mitigation pro- tocols ranging from
traffic rerouting to issuing public health advisories.
This early detection is made possible by the models’ ability to exploit subtle tempo- ral and spatial trends
invisible to static rules. Consequently, real-world deploy- ments integrating ML analytics with IoT sensor
networks not only increase data trustworthiness but materially improve the speed and effectiveness of urban air
quality management.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 430
www.rsisinternational.org
Challenges and Limitations
Sensor Calibration and Drift
Low-cost sensors such as MQ-135 exhibit long-term drift of 8–15% over 18–24 months due to aging,
environmental exposure, and particulate fouling. Humidity and temperature variations can introduce additional
±5–10% error without compensation algorithms. Mitigation strategies include periodic recalibration, auto-
calibration using reference stations, environmental compensation firmware, protective enclosures, and redundant
sensor deployment. These practices restore accuracy to approximately ±5–8% in field conditions [4].
Data Privacy and Security
IoT air-quality systems require encrypted communication (MQTT over TLS/SSL) to prevent data interception
and spoofing. Weak authentication and unencrypted transmission remain common vulnerabilities. Secure
deployments require authenticated APIs, access control, rate limiting, and multi-factor authentication. Open data
standards are recommended to avoid vendor lock-in and ensure portability. Continuous auditing and penetration
testing are essential for maintaining data integrity and public trust [9][10].
Network Connectivity and Reliability
Connectivity disruptions arise from urban dead zones (affecting ~5–15% of monitored areas), rain attenuation
(signal reduction 10–30%), and power interruptions. Modern systems mitigate these issues using multi-network
redundancy, local buffering, adaptive transmission scheduling, and failover mechanisms to ensure uninterrupted
data continuity [8].
Scalability Economics
Although hardware costs range from $40–$250 per node, large-scale deployment introduces installation ($100–
$300 per node), maintenance (15–25% annually), and cloud/data costs ($0–$50 per node/month). Five-year total
cost of ownership typically ranges from $800–$2,000 per node. Cost optimization strategies include modular
hardware, open-source software, OTA updates, infrastructure sharing, and local maintenance training [9][10].
Future Directions and Recommendations
Hardware
Evolution
Next-Generation Sensors: The past few years have witnessed a surge in innova- tion in air quality sensor
hardware, with particular progress in spectral sensing, AI-enabled edge devices and the miniaturization of
analytical platforms.
Recent advances include spectral sensing, AI-enabled edge devices, and miniaturized spectrometers capable of
multi-pollutant detection with improved selectivity and reduced recalibration needs. Edge AI enables real-time
anomaly detection and localized inference using lightweight models, reducing latency and bandwidth. Wearable
air-quality monitors extend monitoring to personal exposure tracking and epidemiological research.
Energy autonomy is advancing through hybrid power solutions combining solar, thin-film batteries,
thermoelectric and vibration harvesting, enabling long-term autonomous deployment with minimal maintenance
[1][2].
Communication and Connectivity
5G Integration:
The deployment of 5G networks for IoT-based air quality monitoring is rapidly transforming possibilities for
sensor data throughput, system latency and network reliability. 5G technology, with its enhanced data rates—
often exceeding 100 Mbps for IoT modules—and ultra-low latency (as low as 1–20 milliseconds), supports the
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 431
www.rsisinternational.org
transmission of large, real-time datasets from dense sensor arrays and enables true city-wide mobile and
stationary node connectivity. One of the most significant architectural advancements is network slicing, which
allows dedicated “virtual networks” over 5G infrastructure to be optimized for mission-critical environmental
monitoring: for example, slices can be configured for ultra-reliability and low latency in disaster-prone zones,
ensuring air pollu- tion alerts or compliance data arrive instantly and securely, even during periods of network
congestion or emergency response. Additionally, 5G’s support for multi-access edge computing allows real-time
analytics, ML inference and rapid public dashboard updates right at the edge, reducing cloud dependency and
enabling rapid regulatory and health interventions.
Satellite IoT: Non-terrestrial networks—especially new constellations of low Earth orbit (LEO) satellites—are
rapidly closing the “connectivity gap” for remote, rural, or infrastructure-limited regions. Initiatives such as
SpaceX’s Starlink (with >7,000 active satellites) and Amazon’s Project Kuiper (targeting 3,236 satellites,
operational by 2029) now offer global, low-latency broadband with bandwidth and antenna technology designed
to natively support IoT traf- fic at scale. Satellite IoT solutions seamlessly integrate with terrestrial 4G/5G,
LoRaWAN, or WiFi air quality sensors, enabling deployment in deserts, moun- tainous terrain, oceans, or
sparsely populated regions—the “final mile” of smart environmental monitoring. Starlink currently supports
download speeds of 50– 220 Mbps and latency as low as 20–30 ms, while Kuiper aims to deliver sub-100 ms
latency and low-cost terminals.
Global coverage and platform interoperability, when combined with auto- matic failover and cloud integration,
mean that even the most isolated sen- sors can continuously participate in coordinated monitoring, alerting and
cross- border environmental inquiries. As both terrestrial 5G/6G and LEO satellite IoT mature, hybrid
architectures are expected to become the norm for govern- ment, enterprise and NGO-led environmental
monitoring initiatives.[8][9][10]
Artificial Intelligence and Analytics
Federated Learning for Privacy and Scalability: Federated learning (FL) is emerging as a transformative
approach for building privacy-preserving, scalable and decentralized AI across IoT-enabled air quality sensor
networks. Rather than transferring raw sensor data to centralized cloud platforms, FL enables each device or
local edge node to train its own model using local data. Only model updates or weights—without individual
sensor’s raw data—are then aggregated on a global server, radically reducing transmission volumes and
safeguarding sensitive environmental or citizen information. Multiple studies show FL achieves forecast
performance very close to that of centralized deep learning while drastically improving privacy and reducing
risks of data leakage, making it especially attractive for smart city or cross-jurisdictional deployments.
Transfer Learning to New Regions and Sparse Datasets: Classic deep learning models require large
amounts of locally labelled air quality data, posing a challenge for cities or regions with newly deployed or
sparse sensor networks. Transfer learning overcomes this limitation by leveraging pre-trained neural networks
(often LSTM, Bi-GRU, or CNN-LSTM composites) developed in data-rich cities or sites and adapting their
weights or patterns to target do- mains with less historical data. Studies demonstrate that transfer learning can
maintain, or even improve, predictive accuracy for AQI and pollutant levels, enable cross-city forecast
deployment and dramatically reduce both computa- tional cost and required annotation effort, accelerating
scalable model rollout and enabling robust operation in cities still building monitoring infrastructure.
Explainable AI for Stakeholder Trust: The “black box” nature of neu-
ral networks has traditionally hindered adoption of deep ML for decision making in public health and policy.
Explainable AI (XAI) frameworks such as SHAP (SHapley Additive Explanations), LIME (Local Interpretable
Model-Agnostic Explanations) and feature attribution methods are now being integrated with ensemble models
(Random Forest, XGBoost) and neural predictors to trans- parently show which sensor features (e.g., PM2.5,
traffic, humidity) are most influential in each prediction. XAI research for air quality and health risk as- sessment
not only builds trust among policymakers, regulators and citizens, but also enables more precise interventions
and calibration.
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 432
www.rsisinternational.org
Physics-Informed Neural Networks: While data-driven AI models ex- cel at pattern discovery, they often lack
explicit modeling of the known physical and chemical processes that govern pollution formation, transport and
dissi- pation. Recent advances integrate physics-informed neural networks (PINNs, neural ODEs, graph neural
networks, etc.) that fuse domain-specific equations (e.g., advection-diffusion, atmospheric chemistry) with
learned, flexible ML rep- resentations. These “hybrid” frameworks deliver superior performance in open urban
systems, offer better generalization to unseen conditions and can deliver accurate, real-time forecasts even in the
face of unmeasured variables, complex topographies, or variable emission sources.
Together, these four pillars—federated learning, transfer learning, explain- able AI and physics-informed
neural design—significantly expand the scalability, transparency, privacy and operational reliability of IoT-
based air quality predic- tion pipelines, representing the vanguard of AI-enriched smart environmental
monitoring in both academic and real-world deployments.[4][5]
Policy and Governance Integration
Real-Time Enforcement and Automated Fines: IoT-based air quality monitoring networks provide city
authorities and regulators with granular, up- to-the-minute data on emissions, enabling real-time detection of
regulatory breaches and rapid intervention. In leading implementations (for instance, mu- nicipal pilot programs
in Taiwan, India and Europe), these systems have been directly integrated with compliance frameworks so that
pollution exceedances trigger automated workflows: regulated entities (like factories) are immediately flagged,
legal notifications are sent and fines can be generated and enforced without manual audits or multi-week lags.
This automation both improves reg- ulatory reach and deters repeat offenses, as both enforcement and
remediation can be nearly instantaneous.[3]
IoT-Verified Carbon Credit Systems: As carbon markets expand globally, IoT sensors are increasingly used
to provide tamper-proof, real-time veri- fication of emissions for carbon credit generation and trading. Recent
research demonstrates that integrating certified sensors with blockchain and smart con- tract protocols ensures
transparent and accurate accounting of emission reduc- tions, guards against fraud and reduces verification time
by over 80% compared to manual systems. Several pilot projects now employ peer-to-peer carbon credit
marketplaces where farmers, industries, or cities can directly offer verified cred- its based on IoT-monitored
reductions, dramatically increasing the efficiency and trustworthiness of environmental offset markets.
Incentives for Low-Emission Behavior: IoT-based monitoring creates actionable “nudges” by supporting
innovative policy instruments: dynamic tolls for traffic, pollution-adjusted taxes or subsidies, reward points for
low-emission activities and grants for industries or neighborhoods reducing emissions below threshold.
Examples include city-specific programs where transportation or manufacturing emissions are tied to public or
business incentives, stimulating community participation in air quality improvement efforts and reinforcing pos-
itive behavior with measurable, verified benefits.
Public Dashboards for Transparency: Transparency and citizen em- powerment are hallmarks of successful
smart city air quality projects. Publicly accessible dashboards—delivered via web, mobile app, or digital street
displays— enable residents to understand and respond to local air pollution risks in real time. Many smart cities
and open-data initiatives also provide full public ac- cess to raw or aggregated historical air quality data, enabling
citizen science, academic research, urban activism and the development of third-party tools. The positive societal
impact is now well documented: open dashboards in- crease accountability, build public trust in environmental
policy and catalyze environmentally-informed choices in everyday urban life.
Health and Environmental Co-benefits
Epidemiological Links Between IoT Data and Health Outcomes: The integration of IoT-based air quality
monitoring networks has enabled new epi- demiological studies that directly connect environmental exposure
data with population health trends, particularly for respiratory and cardiovascular condi- tions. Real-time, high-
resolution datasets allow researchers and public health officials to correlate spikes in pollutants like PM2.5, NO
2
,
or ozone with hos- pital admissions, asthma exacerbations and chronic illness rates, often with neighborhood-
level granularity. In recent field studies, visual analytic tools— such as the 3D HEPA-filter lung model driven
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 433
www.rsisinternational.org
by IoT sensor streams—have powerfully illustrated the progressive and spatially variable impact of pollution on
lung health, helping inform both policy and individual behavior.
Alerts for Vulnerable Populations: Modern IoT air quality systems can deliver targeted alerts to at-risk groups
(e.g., elderly, children, asthma sufferers) through mobile apps, SMS, or web dashboards whenever local air
quality sur- passes health-based thresholds. These timely notifications empower vulnerable people to adjust
activities in real time—avoiding outdoor exposure, modifying commutes or seeking improved indoor air
filtration—thereby reducing acute health risks.
Occupational Health Protections: In industrial and occupational set- tings, IoT sensor networks provide real-
time, zone-specific exposure monitoring for workers. This capability supports automated detection of hazardous
episodes (e.g., CO, NO
2
, dust exposure) and triggers on-site alarms, ventilation system control, or management
notifications, aiding compliance with safety standards and reducing workplace health incidents. These systems
also create valuable data archives for post-incident forensics and regulatory reporting.[7]
Ecosystem Monitoring and Climate Resilience Planning: Beyond direct human health applications, IoT-
enabled networks underpin ecosystem conservation and climate resilience efforts. By monitoring air, soil and
water quality, wildlife movements and microclimate variations, cities and conservation agencies can proactively
address environmental degradation, understand the local impacts of climate change and optimize interventions
for biodiversity and sustainability. AI-driven forecasting models further enhance preparedness by predicting
adverse events (e.g., smog, heatwaves, wildfire smoke) and supporting targeted, real-time ecological
management—a critical adaptive advantage in an era of intensifying climate stress.
CONCLUSION
The global landscape of IoT-based air pollution monitoring systems reveals a striking convergence between
technological innovation and practical, impact- ful deployment. As demonstrated by evidence from over 20
leading research projects, field trials and urban pilots—spanning environments as varied as In- dia’s dense city
centers, Taiwan’s smart city networks, Bulgaria’s machine learn- ing analytics and community-driven setups in
Africa and South Asia—IoT- enabled monitoring has advanced decisively beyond theoretical promise toward
transforming environmental intelligence and public health.
This review confirms several critical advancements:
• Democratization and reach: The dramatic reduction in cost (with node prices as low as $40–$250) now allows
for unprecedented sensor density and spatial coverage, closing longstanding data gaps in low- and middle-
income regions. Community and city-wide deployments are now feasible, with robust scalability from
neighborhood experiments to national grids.
• Real-time, actionable intelligence: Unlike the delayed, coarse resolution of legacy station-based systems, IoT
architectures enable immediate, hy- perlocal monitoring. Data pipelines from NodeMCU, Raspberry Pi and
edge-enabled sensor networks feed directly into cloud and municipal dash- boards, facilitating rapid response
and empowering both authorities and citizens.
• Machine learning integration and predictive power: The transition from threshold-based triggers to advanced
ML (BDBN, LSTM, federated learn- ing, explainable and physics-informed neural networks) is a major leap.
These approaches not only boost accuracy by 15–25% but also support robust anomaly detection, early
warnings (30–60 min in advance) and dynamic learning in new environments—substantially enhancing
health protection and regulatory readiness.
• Multipurpose societal and policy benefits: IoT-driven systems support im- pacts well beyond environmental
compliance. Documented results include PM2.5 reductions of 20–35%, measurable declines in respiratory
hospital- izations, improved industrial compliance and stronger citizen engagement. IoT data streams now
underpin epidemiological studies, trigger targeted alerts for vulnerable populations, enable automatic fines
and incentives, strengthen climate resilience planning, verify carbon credits and drive new models of
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 434
www.rsisinternational.org
participatory environmental governance.
• New frontiers in technology and governance: Innovations such as spectral and wearable sensors, edge AI,
5G and satellite integration, open dash- boards, blockchain-verified carbon trading and hybrid energy
solutions are moving the field rapidly toward fully autonomous, universally accessible environmental
management systems.
Key limitations and research priorities remain:
• Persistent challenges—sensor drift, calibration, network reliability, cyber-security, data privacy, policy
fragmentation and rising mainte- nance/ownership costs at mass scale—are well-documented in both the
literature and case studies. However, coordinated and interdisciplinary advances in resilient hardware,
adaptive analytics, cross-sector standards and community partnership are repeatedly shown to mitigate these
barriers effectively.
• The move toward federated and transferable learning, explainable models and multi-source data fusion is
especially crucial for building public trust, facilitating calibration across geographies and accelerating
adoption in policy and health systems worldwide.
In essence, the IoT revolution in air pollution monitoring exemplifies the convergence of affordable sensing,
real-time cloud analytics, robust AI and open governance. These systems are not only enabling a more detailed
and timely understanding of air quality dynamics, but are legitimizing a new, responsive model for
environmental protection—one in which governments, industry and the public share both data and decision-
making power.
The future of urban and environmental health is now tied to our ability to sustain, scale and responsibly advance
these technologies. By fostering collabo- rative research, open standards, transparent platforms and empowered
citizen science, the promise of IoT air monitoring can be universally realized—driving cleaner air, healthier
communities and a more resilient, data-driven response to the escalating challenges of pollution and climate
change.
REFERENCES
1.
Hindustan Institute of Technology and Science. (2022). Design and Anal- ysis of IoT Based Air
Monitoring System.
2.
Sathyabama Institute of Science and Technology. (2022–2023). IoT- Air Pollution Monitoring System.
3.
Taiwan Association for Aerosol Research. (2023). Development of IoT Technologies for Air Pollution
Prevention and Improvement. AAQR.
4.
Bulgarian Academy of Sciences. (2023). Reliability Analysis of an IoT-Based Air Pollution Monitoring
System Using Machine Learning Algorithm-BDBN.
5.
Karnati, H. (2023). IoT-Based Air Quality Monitoring System with Ma- chine Learning for Accurate
and Real-time Data Analysis. arXiv:2307.00580.
6.
Gueye, A., Drame, M. S., & Niang, S. A. A. (2024). A Low-Cost IoT- Based Real-Time Pollution
Monitoring System Using ESP8266 NodeMCU. Journal of Environmental Engineering and Science.
7.
Dr. NGP Institute of Technology. (2024). IoT Based Pollution Monitoring System for Effective
Industrial Pollution Monitoring and Control.
8.
IJRTI. (2025). Real-Time Air Quality Monitoring Using IoT Sensors. Vol. 10, Issue 4.
9.
International Research Journal of Modernization in Engineering Technol- ogy and Science (IRJMETS).
(2025). Real-Time Air Pollution Monitoring Using IoT and Cloud Integration. Vol. 7, Issue 4.
10.
STM Journals. (2024). IoT Based Industrial Pollution Monitoring System. Vol. 18, Issue 3.
11.
International Journal of Research Methods in Education and Social Sci- ences (IJPREMS). (2025). IoT
Based Air Pollution Monitoring System. Issue 3.
12.
International Research Journal of Engineering and Technology (IRJET). (2025). Air Pollution
Monitoring System. Vol. 12, Issue 5.
13.
Ahmed Gueye et al. (2024). Low-Cost IoT Deployment in Developing Regions. Senegal
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Page 435
www.rsisinternational.org
Environmental Technology Review.
14.
International Journal of Internet of Things and Web Services. (2024). Real-Time Industrial Emission
Monitoring via IoT. Vol. 8, Issue 2.
15.
Sathyabama Institute NAAC Report. (2023). IoT- Air Pollution Monitor- ing System.
16.
Sathyabama Institute AQAR. (2022–2023). IoT Based Air Pollution Mon- itoring System. B.E-ECE
Batch 19–23.
17.
Review of IoT Systems for Air Quality Measurements Based on LTE/4G and LoRa Communications.
(2024). International Journal of Environ- mental Monitoring, 15(4).
18.
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT).
(2024). IoT Based Air Pollution Monitoring System. Vol. 4, Issue 6.
19.
ScienceDirect. (2022). Air Pollution Monitoring System Using IoT De- vices: Review.
20.
ScienceDirect. (2025). Evolving Trends in Application of Low-Cost Air Quality Sensor Networks.
21.
Nature. (2024). Machine Learning-Driven Framework for Real-Time Air Quality Monitoring and
Prediction.
22.
PMC. (2023). Internet of Things Based Visualisation of Effect of Air Pollution on Health.
23.
IEEE Sensors Journal. (2022). Machine Learning for Air Quality Predic- tion and Data Analysis.
24.
ScienceDirect. (2025). Artificial Intelligence in Environmental Monitor- ing.
25.
ACM Digital Library. (2024). Evaluation of Low-Cost Air Quality Sensor Calibration Models.
26.
SSRN. (2025). Air Quality Prediction and Analysis Using Machine Learn- ing.
27.
Omdia. (2025). Satellite IoT Market Landscape – 2025.
28.
Nature. (2025). Explainable AI Analysis for Smog Rating Prediction.
29.
ScienceDirect. (2024). Blockchain-Secured IoT-Federated Learning for Industrial Air Pollution
Monitoring.
30.
AAQR. (2023). What Influences Low-Cost Sensor Data Calibration? – A Systematic Review.
