Measurement: Sensors 27 (2023) 100806

Contents lists available at ScienceDirect

Measurement: Sensors
journal homepage: www.sciencedirect.com/journal/measurement-sensors

Smart farming using cloud-based Iot data analytics

Anil V. Turukmane a, M. Pradeepa b, *, K Shyam Sunder Reddy c, R. Suganthi d,
Y.Md Riyazuddin e, V.V Satyanarayana Tallapragada f
a
School of Computer Science & Engineering, VIT- AP University, Amaravati, Vijaywada, India
b
School of Information Technology and Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, 632 014, India
c
Department of Information Technology, Vasavi College of Engineering, Hyderabad, India
d
Department of ECE, Panimalar Engineering College, Bangalore Trunk Road, Varadharajapuram, Poonamallee, Chennai, 600123, India
e
Dept of CSE, School of Technology, Gitam University, Hyderabad, India
f
Department of ECE, Mohan Babu University (Erstwhile SreeVidyanikethan Engineering College), Tirupati, Andhra Pradesh, India

A R T I C L E I N F O A B S T R A C T

Keywords: The introduction of cutting-edge technologies like Internet of Things (IoT) detectors and drones, and farm sur­
Internet of things veillance, is transforming the farming sector in the Big Data era. Huge quantities of priceless agridata are
Smart farming generated by IoT systems, and cutting-edge application technologies enable the true collection and analysis of
Cloud data
this information. This technological combination, referred to as “smart farming,” enables different agriculture-
Big data
Analytics
based players to analyze plants in real-time and enhance profitability and efficiency in farm and company ac­
tivities with the least amount of work. Even though several precision agriculture methods have been developed
by academics and businesses, it is sadly never possible to apply those strategies to all farms. A custom, semi-
public big data processing infrastructure serves as the foundation for the majority of these applications.
Throughout this article, we suggest WALLeSMART, a virtualized precision agriculture management system used
in India’s Wallonia. The framework presents a broad framework to handle the difficulties associated with col­
lecting, analyzing, storing, and visualizing extremely huge volumes of information on a factual and batch basis. A
first version has indeed been created and then pushed to the limit on several fields, with impressive outcomes.

1. Introduction agricultural practices today [6]. By utilizing information and commu­

nications technologies, the notion of precision farming is extended to
Food scarcity has traditionally been linked to socioeconomic factors ground crop/farm administration, which improves governance and
such as a growing population. The world’s population has increased by decision-making duties by utilizing contextual, condition, and
3–6 billion in the past fifty years, which has exacerbated the need for location-based services [7].
food [1,2]. According to projections, the world’s population is expected The issues of agricultural output in terms of performance, effects on
to rise by more than 30% through 2050, necessitating a 70% growth in the environment, agriculture, and sustainability, must be met via smart
food production [3]. Food security is characterized by the availability of farming. Given that it improves the capital base and environment pro­
adequate, secure, and nutrient-rich food to all individuals around the tection on which farming relies to provide basic human food demands,
world [4]. Such misunderstandings are exacerbated by environmental organic farming is extremely pertinent and closely related to smart
damage and marine pollution, global warming, socially constructed farming [8].
advancement, government actions, and market volatility. Such uncer­ The complicated, multimodal, or unexpected agriculture environ­
tainty puts pressure on agribusiness, which already contributes 20% of ments must be thoroughly examined and comprehended to meet the
global CO2 emissions, to increase efficiency while reducing its impact on issues that smart farming and sustainable agriculture raise [9]. By
the environment [5]. constantly monitoring and analyzing different elements of the physical
The concept of “smart farming” has emerged as a consequence of the world, the abovementioned developing information devices contribute
assistance that biotech and modern electronic techniques like remotely towards this knowledge by providing enormous amounts of information
sensed, virtualization, and the Internet of Everything were providing for at such an extraordinary speed. Therefore it indicates that large amounts

* Corresponding author.
E-mail address: [email protected] (M. Pradeepa).

https://doi.org/10.1016/j.measen.2023.100806
Received 29 December 2022; Received in revised form 18 March 2023; Accepted 20 May 2023
Available online 22 May 2023
2665-9174/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
A.V. Turukmane et al. Measurement: Sensors 27 (2023) 100806

of information spanning numerous heterogeneous sources must be chores. Big data solutions often need to meet two fundamental criteria
collected, stored, pre-processed, modeled, and analyzed [10]. [21]. Collecting a vast true digital signal from numerous sources and
Agriculture “big data” necessitates significant expenditures in facil­ then immediately analyzing this data to produce conclusions. Relying on
ities for processing and storing, including some that must run very this, three cutting-edge software designs termed Lambda, Kappa, and a
instantly [11]. Therefore, “big data analysis” is the phrase being used to hybrid of these are frequently employed with big information systems.
characterize a new generation of procedures created to ensure higher Fig. 1 illustrates how information is distributed on those sides. In­
information gathering, finding, and/or analysis because then farmers formation is uploaded to a storing region known as the master’s infor­
and related organizations can derive financial benefit from some very mation in the upper layers. Typically, batch views are created from this
massive amounts of a huge amount of information [12]. information once it has been handled in the past using a batch-
Since big data and smart farming are both comparatively recent processing engine like Apache Hadoop. To use a stream processing
ideas, it is to be anticipated that information regarding their potential technology like Apache Storm, information is analyzed in real-time to
uses and effects on R&D is not extensively disseminated [13]. Big Data produce an updated representation in the velocity layer. The findings
implementations could well have reached the “high point of inflated were typically out of time since it takes a long time to process large
expectations,” according to Gartner’s hype cycle, according to some quantities of information in the upper layers. This place is filled by
scholars. However, those around believe that big data implementations accessing much more recent information from the velocity layer. The
have already surpassed that point [14]. This analysis intends to shed servicing layer, where users may submit their inquiries, combines the
light on one of the most recent Big Data applications with a connection representations of the levels below.
to precision agriculture and to highlight the much more significant areas It simply employs the speed level, as seen in Fig. 2, which is one
for further study and improvement. branch. In other words, one parallel programming tool is always used to
process information in real-time. Incremental views were used to save
2. Related works the findings. The information is then sent to the people, who conduct
any necessary inquiries. This modification allows the Kappa design to
All technological and social and economic factors are taken into handle the whole master’s information in accordance with requirements
consideration when assessing the research [15]. The condition in this and even enter actual information concurrently. The single information
field, unfortunately, is evolving quickly, and it will likely be totally processor unit of the Kappa design has the benefit of simplifying the
obsolete then after the study is released [16]. Even though it is expected systems as a whole.
that it would have a longer-lasting influence, our study largely focuses When analyzing both true and archived information, the Lambda and
on the economic impacts Big Data would have on agricultural operations Kappa designs provide support. The working manner is nevertheless
or the entire system along it [17]. Monsanto’s Integrated Agricultural divided. That is, to allow a thorough study of the information, the
Systems system, for example, gathers data on things like soil conditions operator must explicitly combine the findings of the different processing
or insect threats and makes it available to producers [18]. The Climate modes. In his work, he develops the BRAID hybrid processor technology.
Corporation suggests using the Climate FieldView system to compile By introducing communication networks seen between the batching
information from diverse sources in one location and give growers di­ engine and the streaming engine, BRAID integrates the analysis of true
agnostics and practical capabilities [19]. and historical information [22]. This makes it possible to automatically
Several nations were developing strategies that promoted the do thorough evaluations at a manageable expense. The information
application of contemporary farming techniques through early adoption mentions “many ways to execute their proposed architecture.” The work
[20]. A Dutch browser platform called Akkerweb, for instance, serves as of developing an innovative framework enabling big data streaming
a gateway to various data sources like meteorological, unit borders, combined with complicated priority scheduling went in the very same
satellite, and information from a commercial agricultural management direction.
software. It keeps georeferenced information, such as aerial photographs
and soil data. 3.2. WALLeSMART system architecture
In this article, WALLeSMART, a general intelligent agriculture cloud-
based platform that supports administration, academics, and producers To develop our solution, we modified a lambda design based on
in running their farms and reaching educated choices. To accomplish the earlier conversations. The distinction between batches and the serious
above, humans set forth four purposes: researching the most recent big analysis in this situation justifies your decision. As a result, we employ
data processing architectural style, proposing a generic architectural excellent tools for each and every task. Additionally, the Lambda
style for the agricultural industry, developing and deploying such a structure significantly boosts speed while ensuring linear scalability and
remedy for our particular case, utilizing real sets of data from milk fault tolerance despite hardware problems. The batches, velocity, and
production and weather data, and visualizing some of this information serving levels, make up our network implementation’s 3 lambda layers.
through an interactive Web-based application. Tools for our implementations were chosen carefully based on a variety
of factors, including if they were free software, accessible, networked,
3. Proposed system extendable, and tested in real-world applications. We shall outline the
levels of our underpinning system in the parts that follow. The infra­
3.1. Smart farming with big data structure undergoes various processes as information passes through
everything. Initially, there are two basic forms of information(i)weath­
As communication networks, IoT, and cloud computing advance and erstationobservationsand (ii) dairy farms.
are used in smart farming, large data can be produced. Volume, velocity, As an MQTT communication protocol, we employ EMQ X. The in­
diversity, authenticity, and value are what define the “large” component formation will then be filtered by a data integration tool when it reaches
of this phrase. Some businesses also added additional aspects to big data. our platform. This tool is in charge of gathering but instead stashing time
To get the full value from this vast amount of data, management and related so that a stream/batch production customer can ingest them
analysis are required. To achieve that, it is necessary to first import, later. We employ Apache Kafka as an information intake mechanism in
combine, and save raw data within a data lake, a central repository. To our research. Thousands of communications for each minute can
screen, aggregate, and start preparing the information for analysis, big perhaps be processed by something on a large scale. Third, the infra­
data solutions must be utilized. These remedies should be predicated on structure evaluates real-time communications following recording in­
both hard workloads and proper jobs. The final stage involves extracting formation to learn about the information. Last but not least, real-time
information from large databases through monitoring and analytical and batch information was exported to storing systems for preservation

2
A.V. Turukmane et al. Measurement: Sensors 27 (2023) 100806

Fig. 1. Big data analytics architecture.

Fig. 2. Kappa architecture.

or further analysis. For NoSQL and relational tables, respectively, we visualization. Our platform includes web-based organizations. Effective
employ Apache Cassandra and PostgreSQL in their design. with Node. js, a middleware based on the V8 Javascript from Google that
Fig. 3 details the WALLeSMART component implementation. Docker makes it simple to create quick and flexible network applications.
container technologies are used to execute the system on such a multi­ OpenStreetMap is used to display farm information on online maps, and
core Linux server, isolating the application from its own surroundings the Chart. js package is used to render the data into interactive charts.
and guaranteeing that it really functions consistently everywhere. In We offer a Web API to assist researchers in gaining access to fields and
actuality, the platform’s information flow processes include the weather information for additional analysis and modeling. For executing
following: (i) Eleveo and Pameses data acquisition Kafka data ingestion; API requests, we make use of the GraphQL query language. Users are
(ii)EleveoandPamesesdataacquisition (iii) Webinterfacedata visualiza­ empowered to request precisely whatever they need from the system
tion. (v)StormandHadoopdataprocessing according to GraphQL, which offers a detailed summary of the data in
The speed work surfaces WalleSMART manipulation and offers data- the API. We also have a comprehensive overview, of user administra­
driven insights via research and monitoring, including real-time tion, position and authorization administration, batch and real-time

Fig. 3. WALLeSMART system components.

3
A.V. Turukmane et al. Measurement: Sensors 27 (2023) 100806

work planning, authorization credential management for managerial mind that the stated value takes into account the time it takes to read
staff, emails, announcements, and numerous other capabilities in our information from the device, process it, and save it to the storage ca­
programme. pacity. Fig. 4 (d) demonstrates that the runtime is lengthy.

4. Results and discussion 4.2. Cloud service orchestration layer

The region of India-Dairy farming industry is the target market for Cloud service lifecycle management is offered dynamically and
our platform, as previously said. The goal is to make it easier for farmers depending on policies by the de développement layer. The existing
to monitor historical observations and weather forecasts and to make Cloud programs and facilities were dispersed, just like the management
them better manage their dairy cows in terms of things like healthcare, capability. With the research carried out and elements, administering
breeding, dairy output, and mobility. This data collection consists of applications over a large set of Cloud locations with such a broad range
55,940 measurement sets with 23 parameters from 30 dairy farms and of features was made possible. To store resources meta-data and policies
7139 different cows. Data from 45 agrometeorological stations that span that benefit high transfer rate updating and retrieving, a networked,
the entirety of Wallonia are included in the Pameseb data set. The his­ permanent repository should be used [23]. a flexible communication bus
torical measurement period is from January 1, 2008, to March 21, 2022, that carries control packets for resource management and operational
while the forecasted data covers the next seven days. It is characterized orchestration. a decentralized policy system that enables localized
by five factors, including sunshine, wind speed, temperature, precipi­ compliance and also has a unified, broad perspective.
tation, and humidity; more than three million previous readings. One Cloudlet computer programme runs on each node of the cloud
platform as part of both the cloud-based orchestrating architecture and
The actual computer and applications installed on it are monitored for
4.1. Performance evaluation
health and configuration by the Cloudlet agent using abstraction layer
APIs. This data is processed globally after being regionally examined and
To evaluate the efficacy of our system, we give several exercises and
sent to the storage devices. Additionally, Cluster is in charge of man­
outcomes. The speed layer’s real-time calculations of the mean,
aging managerial activities, including trying to stand up and down host
maximum, and lowest daily temperatures for the upcoming 7 days are
OSes, and application containers, deploying and decommissioning case
displayed in Fig. 4 (a). It is required to establish forecasts on a frequent
stories, etc. As a result, cloud services interact with a variety of entities
basis because we only receive logical information on an hourly basis.
on a cloud node, starting with the physical server, virtualization, host
The data stream entering Kafka topics is immediately transferred to
OSes, service containers, and service instances [24]. Every one of those
Storm, where the calculations are carried out through Storm bolts, the
organizations performs the required reporting and monitoring duties for
conclusions are sent back to Kafka, and finally, the data is kept in Cas­
programmes;
sandra. Fig. 4 (b), on either hand, shows the outcomes of a batch’s task
Although difficult to create, a distributed system is perfect for
completion that was started more by the batch layer. Determining the
improving the fault tolerance and flexibility of the cloud. Information
average daily, highest, and lowest temperatures are the work at hand.
can be stored and retrieved from a distributed system more quickly than
In a different series of tests, we assessed how well the design per­
from a centralized one. Software and the essential context for Cloud
formed when running batches and real-time workloads. Fig. 4 (c) shows
service orchestration are also both stored in a database.
the results of a month’s worth of execution time measurements for the
same preceding real-time job. Since we receive the very same quantity of
information every time, the estimated duration is often short. Keep in

Fig. 4. Scheduled weather station (a) Temperature forecasting (b) Temperature history (c) Weather data execution time (d) Dairy data over one month.

4
A.V. Turukmane et al. Measurement: Sensors 27 (2023) 100806

4.3. Policy-based service contact the policies repository to determine the devices, applications,
and complete network policies. The orchestration features are distrib­
The orchestrating framework offers plan resource forwarding by uted all through the public cloud such that each cloudlet contains the
directing an incoming support request towards the delivering service logic shown in Fig. 5. For globally disparate installations, dispersed
instances that conform with both relevant business regulations. The control offers higher resilience and flexibility, and quicker orchestra­
organization gets assistance in accomplishing this from the strategy tion. Agriculture is the industry that has the greatest impact on everyone
manager. In this section, the regulatory coordination workflow as on the globe. Understanding the intricate agricultural ecosystems is
described. Managers interact with both the orchestration platform and a more important than ever because of the growing difficulties facing
panel, which offers a simple user interface. Managers can manage and agricultural production.
monitor the cloud platform through this UI, and set up organizational
policies. The policy elements in the UI are being changed by the 5. Conclusions
administrator to suit their requirements. The architecture was adaptable
and might support a range of policies. We launched WALLeSMART, a brand-new cloud-based platform for
A distributed policy database receives business strategies, which are controlling smart farming, in this article. We investigate the much more
supplied through the user interface. The arrival of a work order initiates significant cutting-edge technologies and big data architectural efforts
compliance with the policies. The policy manager pulls pertinent regu­ before outlining the proposed solution. A universal big data framework
lations from the policies repository, like those that deal with businesses, for smart farming was introduced as a result of this. To address the issues
subscriptions, renters, etc. Additionally, the policy administrator pulls of real-time big data gathering, computing, storage, and visualization,
meta information about running and extending the functionality from the following assertions are a Lambda architectural methodology. The
the resources registry. The policy administrator searches throughout platform’s backend displayed strong information systems, whereas the
these two chunks of information for just one active duty example that front earned customers’ satisfaction with batches and real information
complies with both policy constraints before sending the support ticket processing. Furthermore, the study has addressed instances of batching
toward that example. A new sample must always be generated if there and real-time job performance. The platform will need to be expanded to
isn’t one already. To find a specific set of cloud nodes for whom the even more farms, and the user interface will need to be improved with
capabilities fit the policy restrictions when initializing the new service, new capabilities including time information analysis, real-time cow
the capability engine was called upon by the policy manager. The placement, spatial-temporal information, and safety assurances.
product lifecycle manager who provides the application on a cloud de­
vice receives this ranked list from the capabilities generator.
When establishing the new object, the life cycle administrator can

Fig. 5. Policy-based orchestration framework.

5
A.V. Turukmane et al. Measurement: Sensors 27 (2023) 100806

CRediT authorship contribution statement [9] T. Anand, S. Sinha, M. Mandal, V. Chamola, F.R. Yu, AgriSegNet: deep aerial
semantic segmentation framework for IoT-assisted precision agriculture, IEEE
Sensor. J. 21 (16) (2021) 17581–17590.
Anil V. Turukmane: Writing – original draft. M. Pradeepa: [10] R. Akhter, S.A. Sofi, Precision agriculture using IoT data analytics and machine
Conceptualization. K Shyam Sunder Reddy: Validation. R. Suganthi: learning, Journal of King Saud University-Computer and Information Sciences 34
Methodology. Y.Md Riyazuddin: Supervision, Writing – review & (8) (2022) 5602–5618.
[11] B.J. Chelliah, T.P. Latchoumi, A. Senthilselvi, Analysis of demand forecasting of
editing. V.V Satyanarayana Tallapragada: Data curation. agriculture using machine learning algorithm, Environ. Dev. Sustain. (2022) 1–17.
[12] U.F. Mustapha, A.W. Alhassan, D.N. Jiang, G.L. Li, Sustainable aquaculture
Declaration of competing interest development: a review on the roles of cloud computing, internet of things and
artificial intelligence (CIA), Rev. Aquacult. 13 (4) (2021) 2076–2091.
[13] K. Balamurugan, T.P. Latchoumi, T.P. Ezhilarasi, Wearables to improve efficiency,
The authors declare that they have no known competing for finan­ productivity, and safety of operations, in: Smart Manufacturing Technologies for
cialinterestsor personal relationships that could have appeared to in­ Industry 4.0, CRC Press, 2022, pp. 75–90.
[14] S.V.K.R. Rajeswari, V. Ponnusamy, AI-Based IoT analytics on the cloud for diabetic
fluence the work reported in this paper. data management system, in: Integrating AI in IoT Analytics on the Cloud for
Healthcare Applications, IGI Global, 2022, pp. 143–161.
Data availability [15] S.I. Hlatshwayo, M. Ngidi, T. Ojo, A.T. Modi, T. Mabhaudhi, R. Slotow, A typology
of the level of market participation among smallholder farmers in South Africa:
limpopo and Mpumalanga Provinces, Sustainability 13 (14) (2021) 7699.
No data was used for the research described in the article. [16] T.P. Latchoumi, R. Swathi, P. Vidyasri, K. Balamurugan, Develop new algorithm to
improve safety on WMSN in health disease monitoring, in: 2022 International
Mobile and Embedded Technology Conference (MECON), IEEE, 2022, March,
References
pp. 357–362.
[17] S. Karad, R. Thakur, Efficient monitoring and control of wind energy conversion
[1] A. Al-Ansi, A.M. Al-Ansi, A. Muthanna, I.A. Elgendy, A. Koucheryavy, Survey on systems using Internet of things (IoT): a comprehensive review, Environ. Dev.
intelligence edge computing in 6G: characteristics, challenges, potential use cases, Sustain. 23 (10) (2021) 14197–14214.
and market drivers, Future Internet 13 (5) (2021) 118. [18] M. Raj, S. Gupta, V. Chamola, A. Elhence, T. Garg, M. Atiquzzaman, D. Niyato,
[2] M. Monica, P. Sivakumar, S.J. Isac, K. Ranjitha, PMSG based WECS: control A survey on the role of Internet of Things for adopting and promoting Agriculture
techniques, MPPT methods and control strategies for standalone battery integrated 4.0, J. Netw. Comput. Appl. 187 (2021), 103107.
system, in: AIP Conference Proceedings, vol. 2405, AIP Publishing LLC, 2022, [19] G. Szabó-Szentgróti, B. Végvári, J. Varga, Impact of Industry 4.0 and digitization
April, 040013. No. 1. on labor market for 2030-verification of Keynes’ prediction, Sustainability 13 (14)
[3] D. Saha, G.N.R. Devi, S. Ponnusamy, J. Pandit, S. Jaiswal, P.K. Bhuyan, Application (2021) 7703.
of nanotechnology in neural growth support system, in: 2022 IEEE 2nd Mysore Sub [20] S.A. Bhat, N.F. Huang, I.B. Sofi, M. Sultan, Agriculture-food supply chain
Section International Conference (MysuruCon), IEEE, 2022, October, pp. 1–6. management based on blockchain and IoT: a narrative on enterprise blockchain
[4] B. Raviprasad, C.R. Mohan, G.N.R. Devi, R. Pugalenthi, L.C. Manikandan, interoperability, Agriculture 12 (1) (2021) 40.
S. Ponnusamy, Accuracy determination using deep learning technique in cloud- [21] S. Linsner, F. Kuntke, E. Steinbrink, J. Franken, C. Reuter, The role of privacy in
based IoT sensor environment, Measurement: Sensors 24 (2022), 100459. digitalization-analyzing perspectives of German farmers, Proc. Priv. Enhancing
[5] Y. Su, X. Wang, Innovation of agricultural economic management in the process of Technol. 2021 (3) (2021) 334–350.
constructing smart agriculture by big data, Sustainable Computing: Informatics and [22] N. Aggarwal, D. Singh, Technology assisted farming: implications of IoT and AI, in:
Systems 31 (2021), 100579. IOP Conference Series: Materials Science and Engineering, vol. 1022, IOP
[6] B.B. Sinha, R. Dhanalakshmi, Recent advancements and challenges of Internet of Publishing, 2021, 012080. No. 1.
Things in smart agriculture: a survey, Future Generat. Comput. Syst. 126 (2022) [23] G. Vitali, M. Francia, M. Golfarelli, M. Canavari, Crop management with the iot: an
169–184. interdisciplinary survey, Agronomy 11 (1) (2021) 181.
[7] P. Garikapati, K. Balamurugan, T.P. Latchoumi, K-means partitioning approach to [24] S.O. Araújo, R.S. Peres, J. Barata, F. Lidon, J.C. Ramalho, Characterising the
predict the error observations in small datasets, Int. J. Comput. Aided Eng. agriculture 4.0 landscape—emerging trends, challenges and opportunities,
Technol. 17 (4) (2022) 412–430. Agronomy 11 (4) (2021) 667.
[8] M.E. Pérez-Pons, R.S. Alonso, Ó. García, G. Marreiros, J.M. Corchado, Deep Q-
Learning and preference based multi-agent system for sustainable agricultural
market, Sensors 21 (16) (2021) 5276.