Using AI to Improve Sensor Data Analysis for Environmental

Monitoring

Ka Hang Lu

Po Leung Kuk Ngan Po Ling College1Po Leung Kuk Ngan, 26 Sung On Street, To Kwa Wan, Kowloon, Hong Kong,

999077, China

Keywords: Environmental Monitoring, Sensor Calibration, Machine Learning, Air Quality Prediction.

Abstract: Environmental monitoring plays a crucial role in detecting and resolving issues such as air and water pollution,

and climate change. However, the sheer amount of data from environmental sensors produces problems

concerning noise, precision, and real-time processing. The conventional methodologies are not capable of

processing such information efficiently and meaningfully. The work explores integrating AI paradigms with

environmental sensor networks to enhance the value and quality of data processing. The study details various

types of environmental sensors, e.g., air quality, water quality, and multi-parameter sensing modules. It

deliberates on how AI methodologies—such as machine learning, deep learning, and AIoT—may be employed

to filter, decode, and predict sensor information results. The work provides several real-world use cases to

demonstrate how AI enhances environmental monitoring networks concerning accuracy, scalability, and

advance knowledge deliverability. The work sets forth the hope of AI-based answers to transform

environmental sensing to render it more intelligent, agile, and adaptive. The work attempts to offer actionable

knowledge to researchers, developers, and policy-planners designing the monitoring infrastructure of the

decade ahead.

1 INTRODUCTION

Damage to the environment due to pollution,

overexploitation, and climate variability has placed

environmental monitoring at the global forefront.

Modern monitoring networks nowadays heavily rely

on the widespread use of sensor networks that

monitor the quality of air and water, temperature,

humidity, among numerous other environmental

parameters. Such sensors generate massive datasets

that should be quickly and accurately deciphered to

guide decision-making and public protection actions.

Nevertheless, readings from raw sensors are prone to

such issues as noise, data drift, and fluctuations

within sampling rates. Such shortcomings largely

hinder the effectiveness of traditional statistical or

rule-based analysis methods. For example, cheap

particulate matter sensors, which have been installed

to monitor the quality of air, have been reported to

require complicated calibration to achieve precision

during diverse environmental conditions (Yaqoob,

Kumar and Chaudhry, 2024).

https://orcid.org/0009-0008-3581-0579

Artificial Intelligence (AI) offers viable

technology to manage such problems. Through

techniques such as machine learning and deep

learning, AI programs have the ability to identify

hidden patterns within sensor data, identify, in real

time, anomalies, and draw accurate conclusions. AI

has, indeed, been successfully integrated into smart

aquacultures, where data from water quality

automatically dictates optimal settings within aquatic

life (Hu, Chen and Wang etal, 2023).

The integration of AI and environmental

monitoring technologies not only improves accuracy

but also enables scalable, low-latency, and context-

aware decision-making. For smart cities, AI-powered

platforms are increasingly being utilized to monitor

air pollution by reacting to real-time traffic and

weather data (Vishwakarma, 2024).

This work intends to elaborate on how AI

enhances the analysis of sensor data in environmental

monitoring. The work aims to address the following

areas: monitoring water quality, detection of airborne

contamination, and novel AI-powered sensor

Lu, K. H.

Using AI to Improve Sensor Data Analysis for Environmental Monitoring.

DOI: 10.5220/0014362700004718

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2025), pages 551-555

ISBN: 978-989-758-792-4

551

systems. The work explains widely used sensors, AI

techniques that are applicable to environmental data,

and offers case analyses that illustrate the impact of

AI-enhanced analysis. The work concludes by

highlighting current deficits and lines of research to

follow in the future.

2 SENSORS IN

ENVIRONMENTAL

MONITORING

Environmental sensors are fundamental components

of modern-day environmental monitoring tools. They

are deployed in various places—ranging from cities to

agricultural settings—in order to detect variables such

as air quality, water quality, temperature, humidity,

and sound level. The sensors provide crucial real-time

data to detect pollution, track ecosystems, and ensure

public health.

The particulate matter or PM2.5 and PM10

measuring air quality sensors are increasingly used on

urban-scale pollution control networks. While cheap

optical sensors are common because they are cheap,

they are subject to issues related to data accuracy and

calibration whenever they are exposed to changing

weather and humidity conditions. Sensor noise and

drift are long-term problems that interfere with the

precision of the readings, particularly in densely

populated environments (Aula, Lagerspetz and

Nurmi et al, 2022).

For the water quality field, pH, dissolved oxygen,

turbidity, and temperature sensors are typical ones

that receive widespread use. They are essential to

drinking water supply monitoring, aquaculture, and

effluent treatment plants. The key problem within this

field is that the conditions inside the water are highly

variable and nonlinear and may change very

significantly over time. This makes multi-sensor data

interpretation, on a real-time basis, crucial and

complex (Hu, Chen and Wang et al, 2022).

The humidity and temperature sensors are often

incorporated into broader environmental networks,

which allow such uses as indoor air quality control

and agricultural optimization. Although they are

typically stable and inexpensive, they may lose their

accuracy in exceptional or dynamic environments.

Still, when paired alongside processing tools, even

minimum-level sensors provide insightful

information, such as recognizing initial signs of heat

distress or approximating mold growth.

As the scale of monitoring enlarges, many

systems are increasingly resorting to wireless sensor

networks to enable distributed data gathering. These

are the cornerstones of environmental Internet of

Things (IoT) deployments, enabling real-time sensing

over extensive areas. Yet, problems such as data

syncing, power control, and latency still abound. To

counteract them, researchers have increasingly

looked to adopt edge computing and embedded AI so

that processing takes place locally on the device

rather than on centralized servers. As one example,

Borah et al. (Borah, Khanal and Sundaravadivel,

2024). reviewed the adoption of edge computing

within environmental monitoring, reporting uses

within smart agriculture, unmanned aerial vehicles

(UAVs), and underwater robotics. They reported

viable implementation scenarios such as a Raspberry

Pi–based implementation of an edge node which was

capable not only of predicting the yield on cherry

tomatoes but by using minimal cloud data traffic and

using Arduino boards and the IBM Watson IoT

platform to implement an infrastructure to analyze

pollution monitoring. These implementation

scenarios demonstrated not only significant

reductions in latency and energy use, but improved

scalability to achieve real-time monitoring of the

environment.

While they have flaws, environmental sensors still

make up a core component of ecological monitoring.

They are considerably more valuable when combined

with smart data processing technology, and they will

centrally determine the future of sustainable

monitoring systems.

3 AI TECHNIQUES FOR SENSOR

DATA ANALYSIS

AI has become a transforming force in the field of

environmental monitoring. Common approaches to

traditional data analysis often cannot keep up with the

velocity, volume, and variability of sensor data. AI

offers a set of adaptive, data-oriented tools that are

able to learn complex patterns, detect subtle

anomalies, and make accurate predictions within

dynamic environments.

Statistical learning techniques such as decision

trees, random forests, and support vector machines

have been widely applied to sensor data to carry out

tasks such prediction of level of pollution, anomaly

detection, and maximization of resource utilization.

The algorithms are optimal when datasets are labelled

and intervariable relationships are nonlinear. For

example, AI techniques have been considerably

applied to estimate states of water quality from a

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

552

number of inputs using sensors and thereby obtain

automated contamination event early warnings

(Suchetana, Srivastava and Gupta etal, 2023).

Apart from classic algorithms, deep learning

methods have gained widespread preference since

they can handle unstructured and high-dimensional

information. Long Short-Term Memory (LSTM)

networks, among others, prove particularly beneficial

while working with time-series information collected

from environmental sensors, e.g., approximating

dissolved oxygen or detecting sudden changes in air

quality. LSTM is a type of recurrent neural network

(RNN) that was designed to identify long-range

temporal connections within sequential information

by retaining a memory cell and applying gating

functions to control information flow. Such

architecture makes LSTM highly applicable to

environmental monitoring tasks, where the values

from the sensors tend to exhibit temporal patterns,

trends, and oscillations over time. For instance,

LSTM-based deep learning models have been

successfully incorporated into real-time control loops

to monitor and regulate feeding, aeration, and water

treatment in aquaculture programs (De Vita, Mellone

and Di Luccio etal, 2022).

The most recent developments in AIoT—Artificial

Intelligence of Things — have further enhanced the

capability of sensor-based monitoring systems. Edge

devices are directly integrated with lightweight AI

models such that local inference and processing occur

on data, thereby reducing latency and dependencies

on cloud infrastructure. This finds specific

significance in backcountry environmental

application scenarios where connections are weak or

where responses are latency-sensitive. Edge-level AI

enjoys additional advantages, such as being more

scalable and energy-efficient, and therefore viable for

large-scale environmental installations.

Ensemble and hybrid approaches are another

promising direction. Such systems include employing

multiple algorithms or integrating statistical and

machine learning to increase robustness and

generalize better. For environmental applications,

hybrid approaches have the capability to account for

sensor noise, address missing values, and improve the

accuracy of prediction using domain knowledge and

data-driven inference.

Overall, AI techniques are not only enhancing

environmental data interpretation, but they are also

helping to shift from reactive to proactive monitoring.

By translating raw sensor data into usable

information, AI enables more informed decision-

making within environmental protection and

sustainability efforts.

4 CASE STUDIES

The arrival of AI has deeply affected environmental

monitoring by enhancing the efficiency and accuracy

of sensor-based systems. The following section gives

an account of three sets of case studies—relating to

water, air, and novel smart sensor uses—representing

field installations where AI played a crucial role in

improving data interpretation and response

mechanisms.

4.1 Water Quality Monitoring

The AI techniques have effectively been utilized to

improve the monitoring systems of the quality of

urban and ruralwaters. For example, AIQUAM was

developed as an AI model to predict contamination

intensities from the latest inputs by sensors on a real-

time scale. The system utilizes machine learning

algorithms to detect anomalies in turbidity and pH

readings and thereby enables rapid identification of

threats to public health (Priyadarshini, Poojitha and

Vinay etal, 2023).

Hawari et al. proposed, in another contribution, a

data-led framework that combined AI predictive

models and policy-planning tools for the environment

to manage sustainable use of water in Malaysia. The

framework utilized regression and clustering models

on time-series data on water utilization and predictive

outputs within the software on WEAP (Water

Evaluation and Planning) and predicted policy

impacts under various climate and utilization

scenarios. The experimental results showed that the

hybrid model improved prediction accuracy by 15%

compared to classical approaches and enabled

decision-makers to visualize such results as timelines

on reservoir depletion and mismatches between

supply and demand, subject to various policy

constraints (Hawari, Mokhtar and Sarang, 2022).

The AQUASENSE system, which was introduced

by Tran et al. (Hawari, Mokhtar and Sarang, 2022),

integrated low-cost Arduino UNO R3 electronics and

a number of sensors to detect air quality indices such

as PM2.5, CO2, CO, temperature, and humidity.

Machine learning algorithms such as K-Nearest

Neighbour (KNN), Expectation-Maximization (EM),

and Long Short-Term Memory (LSTM) networks

were applied to address missing data and estimate

pollution concentrations. The system managed to

achieve 96% accuracy in real-time one-hour ahead

prediction of air quality indices while utilizing a web

interface to display trends and alarms, thus validating

the system's ability to provide inexpensive, precise,

Using AI to Improve Sensor Data Analysis for Environmental Monitoring

553

and up-to-date environmental monitoring using

inexpensive components.

An IoT platform that can be applied to river basins

within Malaysia uses a combination of remote

networks and AI filtering techniques to continuously

monitor environmental standard conformity on a real-

time basis. Besides offering real-time responsiveness,

this facilitates proactive mitigation strategies (Rollo

and Po, 2021).

For aquacultural purposes, AIoT technology has

been used to automatically control the environment.

The platforms integrate the data from the sensors

using deep learning models to attain optimal feeding

cycle and optimal water quality, resulting in

improved productivity and reduced mortality among

the fishes.

4.2 Air Quality Monitoring

The AI methods have been widely applied to airborne

quality monitoring to enhance the accuracy of the

sensor and enable real-time forecasting. Sarwar et al.

(López-Ramírez and Aragón-Zavala, 2023)

calibrated 16 low-priced PM2.5 sensors in Lahore

using ensemble models such as XGBoost and

Random Forest. The actual calibration enhanced R²

from 0.42 to 0.88 and reduced MAE by over 30%,

enhancing the accuracy of the sensor within urban

areas by a significant amount.

Arifin et al., in yet another study, compared the

use of machine learning algorithms to calibrate

sensors in urban cities located in Malaysia. Nonlinear

methods, like support vector regression and artificial

neural networks, outperformed linear methods, like

multiple regression, to provide improved

performance. The ANN method indicated the highest

accuracy, R² and RMSE values being 0.90 and 3.8

µg/m³, respectively, and was found to better adapt to

noise within the environmentPopescu, Mansoor and

Wani etal, 2024).

The second team designed an IoT-based decision

tree model of a sensor array to predict air quality. The

system effectively forecasted short-term air pollution

occurrences within 90% and 5 5-second processing

time, issuing timely warnings to urban

administration.

Seoul's smart city project linked readings of

pollution to traffic and industry emissions to find

sources of emissions. The system was 89% accurate

in attribution and triggered automatic actions such as

ventilation adjustment and traffic routing, closing a

loop between monitoring and control.

4.3 Smart Sensors in Other Environmental

Domains

Aside from monitoring atmospheres and water, AI

techniques have been incorporated within general-

purpose environmental monitoring networks. A

whole smart monitoring framework was envisaged

within industry regions, integrating noise,

temperature, and particulate sensors and an AI-based

predictive engine. The system operates in real-time

and has the flexibility to offer early warnings on

perilous circumstances from learned patterns (Zhang,

Shu and Rajkumar, 2021). Such use cases highlight

the rising flexibility of AI-based sensor systems,

especially their capability to allow proactive control

within dynamic or unknown environments.

5 CHALLENGES AND FUTURE

DIRECTIONS

Although there has been significant progress, the

marriage between AI and environmental sensor

networks still faces several large issues. Lower-

budget sensors yield bogus or noisy data, which

reduces the AI prediction performance unless suitable

preprocessing and model compensation are

performed. Deployment of AI algorithms in field

monitoring situations, particularly on edge devices,

still proves to be difficult due to insufficient power,

communication, and processing capacity. The

aforementioned shortcomings diminish real-time

response and scalability.

Also, AI models that are trained on one

environment carry a tendency to generalize to other

regions or states, especially when environmental

factors vary unpredictably. On an industrial

application, AI-assisted prediction on air pollution

exhibited very good promise, but experienced

adaptation difficulties within dynamic working

conditions (Ramadan, Ali and Khoo etal, 2024).

Future work could involve enhancing model

robustness, developing adaptive learning

architectures, and extending multi-sensor information

integration to improve performance in large-scale,

diverse environmental installations.

6 CONCLUSION

The article surveyed how AI enhances the processing

of sensor data when applied to environmental

monitoring. Focusing on applied uses on water

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

554

quality, air pollution, and multi-sensory platforms,

the study depicted that AI-based approaches improve

data accuracy, anomaly detection, and real-time

responsiveness by a significant amount when

contrasted with classic approaches.

The subsequent case studies detail the many uses that

AI has been applied to support automated

decisionmaking on an environmental level. From

smart cities to aquaculture, the automated processing

of sensor data has assisted in speeding intervention,

enhancing resource use, and raising system

flexibility.

Through systematic research and practical

implementation, this work confirms that AI not only

provides an efficient tool to monitor the environment,

but an accelerant to create greener and smarter

monitoring infrastructure. The integration of AI into

the environmental system marks a crucial milestone

in enhancing global collaboration on ecological

protection and technological innovations.

REFERENCES

Aula, K., Lagerspetz, E., Nurmi, P., & Tarkoma, S. (2022).

Evaluation of low-cost air quality sensor calibration

models. ACM Transactions on Sensor Networks

(TOSN), 18(4), 1–32.

Borah, S. S., Khanal, A., & Sundaravadivel, P. (2024).

Emerging Technologies for Automation in

Environmental Sensing. Applied Sciences, 14(8), 3531.

De Vita, C. G., Mellone, G., Di Luccio, D., Kosta, S.,

Ciaramella, A., & Montella, R. (2022, October).

AIQUAM: Artificial intelligence-based water quality

model. In 2022 IEEE 18th International Conference on

e-Science (e-Science) (pp. 401–402). IEEE.

Hawari, H. F. B., Mokhtar, M. N. S. B., & Sarang, S. (2022).

Development of real-time internet of things (IoT) based

water quality monitoring system. In International

Conference on Artificial Intelligence for Smart

Community: AISC 2020 (pp. 443–454). Springer,

Singapore.

Hu, W. C., Chen, L. B., Wang, B. H., Li, G. W., & Huang,

X. R. (2022). An AIoT-based water quality inspection

system for intelligent aquaculture. In 2022 IEEE 11th

Global Conference on Consumer Electronics (GCCE)

(pp. 551–552). IEEE.

Hu, W. C., Chen, L. B., Wang, B. H., Li, G. W., & Huang,

X. R. (2023). Design and implementation of a full-time

artificial intelligence of things-based water quality

inspection and prediction system for intelligent

aquaculture. IEEE Sensors Journal, 24(3), 3811–3821.

López-Ramírez, G. A., & Aragón-Zavala, A. (2023).

Wireless sensor networks for water quality monitoring:

A comprehensive review. IEEE Access, 11, 95120–

95142.

Popescu, S. M., Mansoor, S., Wani, O. A., Kumar, S. S.,

Sharma, V., Sharma, A., ... & Chung, Y. S. (2024).

Artificial intelligence and IoT driven technologies for

environmental pollution monitoring and management.

Frontiers in Environmental Science, 12, 1336088.

Priyadarshini, S. H., Poojitha, S., Vinay, K. V., & VA, A.

D. (2023, October). AQUASENSE: Sensor Based

Water Quality Monitoring Device. In 2023

International Conference on Self Sustainable Artificial

Intelligence Systems (ICSSAS) (pp. 1786–1789).

Ramadan, M. N., Ali, M. A., Khoo, S. Y., Alkhedher, M.,

& Alherbawi, M. (2024). Real-time IoT-powered AI

system for monitoring and forecasting of air pollution

in industrial environment. Ecotoxicology and

Environmental Safety, 283, 116856.

Rollo, F., & Po, L. (2021). SenseBoard: Sensor monitoring

for air quality experts. In CEUR Workshop Proceedings

(Vol. 2841).

Suchetana, B., Srivastava, B., Gupta, H. P., & Saharia, M.

(2023). Promoting sustainable water usage and

management with water data, AI and Policy. In

Proceedings of the 6th Joint International Conference

on Data Science & Management of Data (10th ACM

IKDD CODS and 28th COMAD) (pp. 308–311).

Tran, Q. A., Dang, Q. H., Le, T., Nguyen, H. T., & Le, T.

D. (2022). Air quality monitoring and forecasting

system using IoT and machine learning techniques. In

2022 6th International Conference on Green

Technology and Sustainable Development (GTSD) (pp.

786–792).

Vishwakarma, S., & Vishwakarma, S. (2024). Application

of AI and IoT technologies to control air pollution in

smart cities. In 8th IET Smart Cities Symposium (SCS

2024) (Vol. 2024, pp. 872–876). IET.

Yaqoob, I., Kumar, V., & Chaudhry, S. A. (2024). Machine

Learning Calibration of Low-Cost Sensor PM2.5 data.

In 2024 IEEE International Symposium on Systems

Engineering (ISSE) (pp. 1–8). IEEE.

Zhang, X., Shu, K., Rajkumar, S. A., & Sivakumar, V.

(2021). Research on deep integration of application of

artificial intelligence in environmental monitoring

system and real economy. Environmental Impact

Assessment Review, 86, 106499.

Using AI to Improve Sensor Data Analysis for Environmental Monitoring

555