Precision agriculture using IoT data analytics

and machine learning

Subalakshmi M
Department of Computer Science and Engineering
Government College of Engineering, Tirunelveli

Abstract
In spite of the insight commonality may have concerning agrarian practice, fact is that
nowadays agricultural science diligence is accurate, precise, data-driven, and vigorous than
ever. The emanation of the technologies based on Internet of Things (IoT) has reformed nearly
each industry like smart city, smart health, smart grid, smart home, including “smart agriculture
or precision agriculture”. Applying machine learning using the IoT data analytics in agricultural
sector will rise new benefits to increase the quantity and quality of production from the crop
fields to meet the increasing food demand. Such world-shattering advancements are rocking
the current agrarian approaches and generating novel and best chances besides a number of
limitations. This paper climaxes the power and capability of computing techniques including
internet of things, wireless sensor networks, data analytics and machine learning in
agriculture. The paper proposed the prediction model of Apple disease in the apple orchards
of Kashmir valley using data analytics and Machine learning in IoT system. Furthermore, a
local survey was conducted to know from the farmers about the trending technologies and
their effect in precision agriculture. Finally, the paper discusses the challenges faced when
incorporating these technologies in the traditional farming approaches.

Keywords
Internet of things (IoT) ; Data analytics (DA) ; Machine learning (ML)

1. Introduction
Agriculture is a crucial part of India’s economy, employing 60-70% of the workforce. Despite
having abundant arable land and water resources, agricultural productivity remains low.
Factors include outdated practices, limited technology, and lack of awareness among farmers.
Pests and diseases also impact crops significantly, decreasing yields and posing risks to food
security. For example, crop yield losses due to pests and diseases can range between 20%
to 40% globally, with similar trends observed in India, especially in semi-arid areas.
Population growth pressures food supply, making it necessary to increase agricultural output.
This can be achieved by implementing smart technologies that use resources efficiently. A
reliable system is needed to monitor crop health, predict diseases, and suggest timely
interventions. Such a system would help control pest-related issues without relying heavily on
chemicals, improving both crop quality and quantity.
The Internet of Things (IoT) offers promising solutions through connected sensors that monitor
soil, weather, and crop conditions in real-time. Wireless sensor networks (WSNs) and Radio-
Frequency Identification (RFID) are integral to IoT, enabling
extensive data collection across fields. IoT-generated data includes temperature, soil
moisture, humidity, and other parameters, which, when analyzed, help predict and prevent
crop issues. The IoT concept, introduced by Kevin Ashton in 1999, connects diverse objects
and devices, allowing real-time access to information.
Precision Agriculture (PA) uses IoT and data analytics to optimize farming practices. For
instance, sensors and drones collect data to detect pests, soil conditions, and crop health,
enabling farmers to apply precise amounts of fertilizers, pesticides, and water only where
needed. Farmers who adopt PA have reported increased yields and return on investment. IoT
applications in PA include crop identification, performance monitoring, and optimized
machinery use.
Data analytics and machine learning are crucial for managing the large, complex datasets
generated by IoT devices. This data includes crop patterns, weather conditions, and soil types.
Tools like MapReduce and Spark can process and analyze this information, revealing insights
that support crop yield predictions and disease management.
This paper aims to:
1. Review the use of IoT and data analytics in precision agriculture.
2. Present a case study of IoT adoption in Kashmir’s apple orchards.
3. Identify challenges traditional farmers face when integrating IoT and data analytics,
grouped into commercial, technical, sectoral, and data-related issues.
The paper is structured as follows:
• Section 2: Literature review on IoT in agriculture, covering sensors, networks, and
data analytics applications.
• Section 3: Case study on IoT application in Kashmir apple orchards.
• Section 4: Challenges in adopting IoT-based precision agriculture.
• Section 5: Conclusions and future directions.

Fig. 1.1. Internet of Things and Data Analytics Relationship.

2. Literature survey
The literature in precision agriculture reveals a wealth of contributions aimed at advancing
agricultural practices through IoT, machine learning, and data analytics. These innovations are
primarily organized into three main categories: 1) IoT Sensors and Networks, 2) Data
Analytics, and 3) Smart Systems Applications in Agriculture.
2.1 IoT - Sensors and Networks
Wireless Sensor Networks (WSNs) are essential in agronomy for environmental and soil
monitoring, providing critical insights into crop health. For instance, WSNs have been
employed to create irrigation schedules based on real-time data on soil moisture, salinity,
temperature, and humidity. Ahmed et al. (2018) introduced a scalable IoT network for
monitoring rural agricultural fields, combining Wi-Fi-based networks and fog computing for
improved energy efficiency and reduced delay. Bhargava et al. (2014) used WSN to develop
a Decision Support System (DSS) for detecting apple scab disease in Himachal Pradesh, and
Muangprathub et al. (2019) developed an IoT system that optimizes water usage for crops like
lemons. Their system includes a control box for data collection, a web application for data
analysis, and a mobile app for notifying farmers about soil conditions, ultimately enhancing
yield by adjusting temperature and humidity levels.
Further research, such as Shinde and Kulkarni (2017), integrates IoT with machine learning
to predict crop diseases by processing environmental sensor data and sending alerts to
farmers. Jawad et al. (2017) reviewed wireless technologies, finding LoRa and ZigBee
particularly effective for precision agriculture due to their range and low power consumption.
Meanwhile, Hsu et al. (2018) proposed a cloud-based IoT platform for large-scale data
analysis, enabling automation in agricultural monitoring and pest management.
2.2 Data Analytics
The shift from manual to data-driven approaches in agriculture has been transformative. Data
analytics, especially machine learning, facilitates the prediction of crop diseases and yields
based on patterns in weather and soil data. Machine learning models such as Support Vector
Machines (SVM), Artificial Neural Networks (ANN), and Logistic Regression have become
prevalent. Singh and Gupta (2018) classified apple diseases using algorithms like SVM and
K-Nearest Neighbor, achieving up to 99.4% accuracy. Geetha (2015) used IoT and machine
learning to predict potato blight by tracking environmental parameters and implementing a
back-propagation network with 94% accuracy. Similarly, Aggelopoulou et al. (2011) employed
image processing to predict apple orchard yields with an accuracy over 80%, linking flower
density with fruit yield.
2.3 Applications of IoT and Data Analytics in Smart Agriculture
IoT-based smart systems empower farmers with real-time data and analytics for improved
decision-making. These systems integrate sensors, mobile apps, and DSS to provide insights
and automate tasks such as irrigation, pest control, and crop monitoring. Systems like
Smartnode by Fonthal et al. (2018) and SEnviro by Trilles Oliver et al. (2019) showcase how
IoT applications can monitor agro-climatic parameters and notify farmers about optimal
treatment times, using edge computing to reduce data transmission. Additionally, Balaji et al.
(2018) and Jayaraman et al. (2016) introduced platforms like SmartFarmNet, which facilitate
data collection on crop conditions, soil fertility, and irrigation needs, helping farmers to
enhance crop productivity.
In summary, advancements in IoT networks, data analytics, and smart systems applications
are revolutionizing precision agriculture, providing farmers with robust tools to improve crop
health, optimize resource usage, and boost yield. This review outlines how these technologies
can be leveraged to mitigate challenges, improve energy efficiency, and increase productivity
in modern agriculture.

Fig. 2.1. Precision Agriculture Model.

 Fig. 2.2. Applications related to precision agriculture.

2.3.1. Soil selection and planning

Soil is fundamental to agriculture, and analyzing its physical, biological, and chemical state
through soil testing helps farmers make informed decisions. Precision farming relies on
efficient soil management to maximize crop yield on limited land. Modern sensor-based
technologies aid in assessing soil nutrients, irrigation levels, and other factors to select the
optimal crops. Tools like Agrocares’ Lab-in-a-Box enable farmers to conduct soil tests without
lab access, enhancing precision in soil and crop selection.
2.3.2. Irrigation of crop fields
Effective irrigation management is essential to conserve the 0.5% of Earth's fresh water
accessible to plants and animals. Agriculture uses 70% of this resource, with even higher rates
in some regions. IoT and WSN technologies improve irrigation efficiency by monitoring soil
moisture and weather data to determine when irrigation is necessary. Systems like the Crop
Water Stress Index (CWSI) assess field moisture needs, integrating sensor and weather data
to guide irrigation, helping conserve water and maintain crop health.
2.3.3. Fertilizers
Fertilizers provide essential nutrients like nitrogen, phosphorus, and potassium (NPK), which
support plant growth. However, excess fertilizer disrupts ecosystems. Precision agriculture
uses technologies like WSN, satellite imagery, and NDVI (Normalized Difference Vegetation
Index) to monitor soil and crop health, guiding the efficient application of fertilizers. Advanced
tools, including Variable Rate Technology (VRT) and geo-mapping, help minimize
environmental impact while maintaining soil health.
2.3.4. Crop disease
Crop diseases, affecting 20-40% of global yields, pose significant financial and ecological
challenges. IoT and wireless sensors, coupled with drones and intelligent devices, monitor
crop health and detect disease, enabling timely intervention. Approaches like chemigation and
vehicle sprays under precision agriculture target disease management more effectively.
2.3.5. Pest management
Inappropriate use of pesticides harms the environment and human health. IoT-based pest
management systems enable real-time monitoring and disease prediction, improving pest
control efficiency while reducing reliance on harmful chemicals.
2.3.6. Yield monitoring
Yield monitoring tracks crops from cultivation to harvest, supporting farmers in optimizing both
yield quality and quantity. Technologies like wireless sensors and optical monitors assess
storage and transport conditions, while environmental monitoring informs decisions. Studies
confirm that crop quality and yield are influenced by factors such as self-pollination and
climatic conditions, helping farmers achieve the best outcomes.

3. Adoption of IoT (Sensors) and Data Analytics in tradtional Kashmir farms for
the prediction of apple scab
Apple is one amongst the most grown fruits in the world. It grows mostly in dry areas. In Jammu
and Kashmir where people from most of the districts are dependent on farming and agriculture,

it produces the large quantity and diverse apples in India. Its industry is worth of thousands of
crores in J & K. As shown in Table 3.1 the district wise production of Apple crop in Kashmir
division.
Table 3.1. District wise production of apples in Kashmir.

But due to the various diseases, wide range of pests and lack of technology in the fields we
lose both quantity and quality of the crop. Apple scab is the most troublesome disease for
apple growers in all parts of Kashmir. Scab is caused by a fungus that infects both leaves and
fruit. Scabby fruit are often unfit for eating and continued infection of leaves weakens the tree.
Small black spots develop on fruit, enlarging more slowly than on leaves. As these spots grow
and become older, the center loses the velvety appearance and becomes brown and scabby.
Heavily infected fruit becomes deformed and cracked when infected at an immature stage.
Apple scab infections occur during wetting periods when moisture stimulates the pathogen
spores to germinate and penetrate plant tissue. For this we did literature survey related to the
crop and the factors affecting its growth. We did a local survey to find data regarding the factors
on which we were going to work on. So, for this purpose we went to SKUAST (Sher-e-Kashmir
University of Agriculture Sciences and Technology, Srinagar, Kashmir) and visited
departments of Soil and Pathology to gather the relevant data. Therein, certain information
regarding the Apple Scab disease was collected. At SKAUST we came to know about the
factors affecting the growth of Apple crop. We analyzed the collected data to verify the factors
which actually affect the production and occurrence of diseases. Studying the R-Squared
values, we see factors like pH, P(Phosphorous) and K(Potassium) play a very important role
in the overall production of crop as shown in Table 3.2 OC(Organic content) is a slightly less
important factor while nitrogen plays the least important role. Based on these values, we
designed some prediction models as shown in Fig. 3.1. Factors like pH, K, P and OC together
are crucial for a quality crop as is evident from its R-Squared value 0.788. The scab prediction
table given here can be used to determine whether or not conditions have been sufficient for
infection so that appropriate spray decisions can be made. Based on a certain combination of
temperature and hours of leaf wetness, we can identify whether that region is safe (no chance
of apple scab occurring) or unsafe (higher probability of the disease occurring with less
incubation days) (MacHardy et al., 1989, Belete and Boyraz, 2017). From the data following
two graphs have been drawn see Fig. 3.2. Safe zone where there is 0% chances of apple
scab while Fig. 3.3 shows the unsafe zone where there is no chance to stop apple scab.

Table 3.2. Relationship between production and factors.

 Fig. 3.1. Plots of Factors vs Production.

Fig. 3.2. Safe zone.

 Fig. 3.3. Unsafe zone.

3.1. Framework

The main aim of this study was to predict the apple scab that is the most common disease of
apple crop. In this problem we used the real time data of wireless sensor/IoT nodes as input
for linear regression model. In this an application is developed for farmers which is simple and
user friendly to inform them about the status of their apple orchards on real time basis. The
framework consists of a number of WSN/IoT nodes scattered in the orchards of the apple with
nearby gateway for collection of the data from the mesh of the nodes While performing the
research on the said problem various steps have come in the way that forms the basis of our
study. These steps are shown in the Fig. 3.4. The nodes are deployed in the apple orchard for
data collection. A network is established between the nodes, data acquisition is done by
nearby gateway or a fog node from prefixed number of nodes in the network. The data analysis
is done on the real time basis for each location/orchard for quick action if needed. In this work
data is collected from many orchards of the Kashmir Valley using the same models as
discussed above. The results are shown in Section 3.6. The detailed framework is explained
below:

3.2. Sensors assembled

In precision agriculture IoT plays an important role. IoT are the embedded systems which have
the capability to sense and share information related to any field in any application. IoT devices
and its services play an important role in wireless world as they can provide users all required
services at anywhere at any time. Diverse number of sensors related to agriculture like plant
sensors, soil sensors, environmental sensors etc. collect the real time data respectively. This
data is very helpful in predict the crop diseases. In our prediction model we required to monitor
two parameters temperature and hours of wetness in a field.The sensors comprised of an IRIS
mote and MTS420 sensor board. The IRIS is a 2.4 GHz module used for enabling low power,
wireless sensor networks. MTS420 offer five basic environmental sensing parameters and an
optional GPS (global positioning system). The five basic environmental sensors included in
MTS420 are temperature, humidity, barometric pressure, ambient light sensor and dual-
axis accelerometer. Among these we are interested in temperature and Leaf wetness hours
only. For this DHT22 sensor is used which is embed on MTS420, gives us temperature and
humidity. Leaf wetness hours are then calculated using humidity.
3.3. Sensor/IoT network setup
This step involves setting up of the WSN for capturing the data. We created a WSN of several
sensors and a gateway. Six sensors have been used to cover the particular area of an orchard.
Each sensor comprises of a IRIS fitted with a MTS 420 sensor board. The gateway is
incorporated in WSN that provides connectivity back to the wired world and distributed nodes
using Zigbee module. The IRIS needs to be programmed to be able to act as a sensor with
sensor board. Six such sensors were developed which are enough to cover the length of an
average apple farm. For the gateway, MIB 520 fitted with an IRIS is used. Here the IRIS is
programmed to act as the gateway in order to communicate with the computer. For
programming of IRIS and MIB 520 MoteConfig Application is used. MoteConfig is a Windows-
based GUI utility for programming Motes. This utility provides an interface for configuring and
downloading pre-compiled XMesh/TinyOS firmware applications onto Motes. MoteConfig
allows us to configure the Mote ID, Group ID, RF channel and RF power. We can also enable
the over-the-air-programming feature present on all XMesh - based firmware. High-power and
low-power XMesh applications are available for each sensor board and platform. The Local
Program tab is used to upload firmware onto the Motes via a gateway. To program motes
correctly, we set up the hardware as follows:
• The gateway(in this case the MIB 520) should be powered and connected to the PC
via a serial, USB or Ethernet port.
• The motes should be firmly attached to the gateway.
• The motes should be turned off before the programming.

3.4. Data acquisition

Next step is to collect the data from various parts of the Kashmir Valley. Data collection is one
of the basic and fundamental task while doing data analysis. To collect data, we selected three
districts of Kashmir: Zakura Area of district Srinagar, Pinpora area of district Kulgam and
Kanelwan area of district Anantnag, spread over a area of 4–5 kanals, 7–8 kanals, and 6–8
kanals respectively. The system was kept in place for over 12 h to collect a single days data.
Care was taken to ensured that the motes weren’t damaged by water leading to wrong
readings. To note any variation sensors were kept 100 meters apart, so that whole data from
the field can be collected. MOTE-VIEW monitoring software was used as an interface between
user and the deployed WSN. It provides the tools that simplifies the deployment and
monitoring of WSN. It helps in data collection into the database. We used PostgreSQL
database for it, this resides on our personal computer system (acts as a localhost), a remote
server.
3.5. Analytics
After the WSN was deployed, the sensors logged their data in a Postgres database. Next step
was to analyse this data and find a prediction model fitting the data captured. The data from
the database table first extracted in the form of a .CSV file.
Data pre-processing: The .CSV file contains certain extra columns which are not required by
us. These columns are dropped first. Next we need to take care of the anomalous values such
as a high peak in temperature which is impractical. This kind of thing happens only few times
but can harm our results. Hence are removed.
Analysis using linear regression: We applied the machine learning on the same data and
predicted the status and requirements of the disease treatment in advance by applying simple
linear regression model. We analysed the captured data and found the prediction model fitting
that data. In our prediction model using linear regression we have three variables. T, W, and I
are respectively the temperature, duration of wetness of leaf and incubation period of the
infection causing pathogen. So we have analyzed the collected data and found a prediction
model fitting the data captured. Our model has three variables as such. The first of these is
the temperature (T), second is the duration of wetness (W) and the final is the incubation
period of the infection causing pathogen (I). The main purpose was to find a relationship which
makes I as a function of W and T. Later on this equation is found to be useful in finding the
vulnerability of some specific combination of wetness duration and temperature with respect
to an attack of infection. In a simple linear regression model, a single response measurement
is related to a single predictor (covariate, regressor) X for each observation.

3.6. Results
Table 3.3 shows the table of results obtained by using linear regression and shows the
Estimated Value, Standard Error, True Value of Intercept, Average Temperature and Wetting
Duration. The Eq. 3.3 thus obtained using linear regression model shown above indicates that
the incubation period is being predicted by the temperature value and the value of wetting
period. With decreasing in the values of temperature T and wetting period W there is an
increase in the value of incubation I. The Algorithm 1 shows the various steps used in order
the get the correct results.

The results showed the incubation period of the areas based on safe and unsafe-zones. The
incubation period for Kulgam was 5 days and for Anantnag district it was 3 days. Hence
Kulgam and Anantnag were categorized as unsafe zones while as Srinagar was categorized
as safer zone under the present climatic conditions. The system was proved to be useful for
the local farmers and hence we got our results attested. In Fig. 3.7 the residuals vs. fitted is a
scatter plot of residuals on the y axis and fitted values (estimated responses) on the x axis.
The scatter plot is used to detect unequal error variances, non-linearity, and outliers. Any data
point that falls directly on the estimated regression line has a residual of 0 i.e. no error in the
predicted value. If we compare the previous work done with the proposed work, wherein they
have not used the real time data for prediction of apple scab disease hence time to respond
to the disease is more but by time we come to know about the disease the response time has
already elapsed and the crop will be affected. In our approach using the IoT for disease
prediction the delay is minimised to the extent that if a corrective mechanism like temperature
control, spray of chemicals, pesticides, watering etc. for real time precise agriculture can be
used. Although we may need to increase the number of parameters for disease prediction in
our proposed mechanism in future but this work can give us a platform, architecture and
algorithm to move ahead.

Fig. 3.7. Residuals vs. fitted.

4. Challenges in IoT Adoption for traditional farmers
Precision agriculture is a powerful technology to meet the challenges of the increasing needs
from scarce resources. Besides the advantages there are number of challenges which
agriculturalists, farmers, researchers and scientists are facing while adopting it. For this
purpose a local survey was conducted to know what the regional people think about it.
4.1. Survey setup
Site Description: Jammu and Kashmir is the northernmost geographical region of the Indian
Subcontinent. Kashmir lies between latitudes 32°and 36°N, and longitudes 74°and 80°E. It
has a different climate for every region owing to the great variation of the level of the altitude.
Wheat, maize and rice crops grown in about 250,000 hectares 210,000 hectares and 110,000
hectares area respectively are the major cereal crops of Jammu and Kashmir division. Apple
is the valuable cash crop of Kashmir Valley. In India, agricultural risks are provoked by variety
of factors, like climatic variability/change, extreme weather events, crop
management practices and soil fertility status etc. which ultimately leads to uncertainties in
yields and prices of the grain. We did the local survey in Kashmir Valley by using precision
agriculture in the crop fields using Information and Communication Technologies (ICT) like
IoT, Machine learning, Data analytics etc. The population of this study are small scale farmers
to large scale farmer using simple technologies and mostly traditional ways in their apple
orchards for agricultural information to improve their productivity. In order to bridge the gap of
knowledge and skill regarding agriculture in this area, we asked some questions through
interview to know its basic causes. 1100 people were selected who were engaged in these
from three districts Kulgam, Srinagar and Anantnag where we have deployed our proposed
system. Random sampling technique was used and 330 respondents (33% from the total
frame) were interviewed in study area based on the population intensity. Table 4.1 shows the
survey site, number of farmers selected in each site and number of respondent in the sample.
While this survey various questions were asked to the farmers. Most of them were
uneducated, some of them were educated, few of them have the knowledge of IoT and other
technologies. In order to know their views various questions were asked to them either through
questionnaire or by interview. Table 4.2 shows the various questions and the percentage of
corresponding responses in terms of Yes or No. The positive (yes) response was mostly
received from those farmers who were either educated, or whose field was under investigation.

4.2. Results
From the last decade, overall amount of funds for the agronomy area has developed enormous
nearly 80%. In accordance with the experts of the agronomy sector smart agriculture is
capable to take on an important role in attaining the proliferating demands of food of the
increasing population throughout the globe. The financial worth of the worldwide smart aggie
marketplace is approximately $4.6 billion with CAGR (Compound Annual Growth Rate)
between 2015 and 2020 being just a touch under 12% (Fakhruddin, 2020) mentioned in a
fresh report. It has been predicted in many foreign countries especially in United States alone
agriculture software will nearly increase around more than 14% between current and 2022.
From the above it is expected that development and propagation of precision farming should
be vigorous, efficient, etc. But like other technology based applications precision agriculture
also faces number of challenges as we have seen in the local survey in Jammu and Kashmir
that there are number of challenges. The challenges faced are mentioned in Fig. 4.1 and are
discussed below:

Fig. 4.1. Challenges of Adopting Precision Agriculture.

4.2.1. Challenges based on commercial

Due to the tremendous advantages of the current technologies like Internet of things, wireless
sensors, actuators, today’s farmers are interested in business models. Using IoT technologies,
they will accumulate the data from their farms to support their revenue. The data provided by
the farmers are exploited by the most of the existing IoT platform service providers. They
provide unrestricted limited services and full services with diverse level of subscriptions. This
becomes the part of argument for farmers to supervise their owned data.
Budget: While developing any solution for smart agriculture, cost analysis should be done. As
we know in IoT, the initial costs are very high due to the variety of sensors, gateways, base
station infrastructure (Farooq et al., 2019). After this maintenance is also an important part of
cost analysis. Here the researcher’s goal should always remain such that increasing the
production while decreasing the cost requirement.
Unawareness of Information and Technology era: As we know most of the agricultural working
is happening in rural areas where farmers are not well educated. They are unable to accept
the new trends and technologies early. This is the reason of slowing IoT in rural areas. So in
 Fig. 4.1. Challenges of Adopting Precision Agriculture.

order to overcome this issue, we need to educate farmers in order to generate best results
(Marjani et al., 2017, Farooq et al., 2019).
Loss of manual employment: Around 60–70% of the workforce are working in the primary
sector globally. The number is high particularly in Asia, Africa and Oceania. As the agricultural
practices become automated and mechanized -a very great percentage of the agronomy
workforce will lose their employment. It will make it necessary to accommodate these employs
in other employable sectors. This will be only possible if these sectors are able to absorb such
a very huge volume of unemployed people. But due to the poverty in most of the developing
countries particularly in India it is hardly possible. So besides the large-scale benefits the
precision agriculture lead but a massive displacement of labors may cause discontent and
disappointment among populace.
Benefits not immediately apparent: While developing any system in any field, QoS is very
important. In big IoT data and machine learning used in precision agriculture QoS should be
highlighted at each layer so that system will give best results at end (Al-Fuqaha et al.,
2015, Huang et al., 2017). There are number of challenges especially while transferring data
from one layer to another QoS is usually compromised. Hence there is a need to do more
research work by providing a solution that will guarantee QoS throughout the network system
at each layer.
4.2.2. Challenges based on sectorial
There are number of parameters that need to considered for distribution of IoT systems. One
of the main factor is the ability of an IoT device to place anywhere for perfect operation and
remain connected globally. This means the ability to connect anywhere at anytime with least
configuration. Other features are ability to support roaming and provide reliability to the system
for proper results. We need to consider the IoT technology infrastructure for the static IoT
devices.
Size of individual management zones: Production function refers to the quantity and quality of
output keeping the land area under cultivation same. But the production function varies for
most of the crops, in different zones of field etc. (Fakhruddin, 2020). But unfortunately farmers
are not aware about it. Precision agriculture cannot be optimized unless and until farmers
become aware about varying production function e.g. excessive or moderate use of fertilizers,
pesticides etc. will lead to crop damage.
Interoperability: It refers to the capability of a device, product, service, system to be operated
with others reciprocally. With the growing Original Equipment Manufacturers (OEM)
approaching with innovative and advanced precision agronomy equipments, devices,
platforms, etc., interoperability becomes the most important point of concern (Fakhruddin,
2020). This involves number of issues related to the technical, semantic, syntactic and
organization interoperability (Elijah et al., 2018, Brewster et al., 2017, Vermesan, 2010, Tayur
and Suchithra, 2017). So in order to overcome this issue researchers should focus on building
the farmer friendly platforms which are holistic instead of smart standalone devices, gateways
or other smart appliance.
Supervisory challenges: Organizing, managing and permissible foundation concerning the
authority and proprietorship of field data between the owners of the fields and the data
companies is required to sort out. Regulations like data privacy, data security, technical issues,
available resources like bandwidth, etc. differ from country to country. Differing these
supervisory features affect the use cases related to the agriculture. Hence due to the
heterogeneity of supervising techniques becomes a challenge for most of the farmers.
4.2.3. Challenges based on technicality
Technology make jobs easy. Every work based on technology becomes automated using IoT
and other mechanized ways. This increases the dependence of health, traffic, home and other
related jobs on the technology. It can be very dangerous at times even if any individual
subsystem of the system will stop working. Hence can become the cause of loss. If we take
the example of the precision agriculture as in our case most of the hardware units are exposed
to the outside environment. This makes it more possible that the sensors may get destroyed
due the various reasons like, rain, wind, snow etc. So in short we can say that we cannot meet
the food safety. Compromising the food safety means system is useless. Sometimes the
downtime can be caused by power failure. Hence we need to make sure we are keeping the
backups for everything in order to meet the requirement. Single point failure can never lead to
success.
Internet Connectivity: In rural areas across the globe particularly in developing countries like
India high speed and reliable internet connectivity is absent (Fakhruddin, 2020). Due to
the wireless communication various issues like multipath propagation, which occurs due to
the buildings trees etc. the IoT system is not able to work consistently. So there is a need to
develop most robust and reliable technologies (Ojha et al., 2015).
Security Factor: It refers when the said system is compromised to restore personal
information. It is one of the most important issues in every field of study. In our case of big IoT
data and machine learning it may compromise any sort of information at any layer, lower or
higher (Marjani et al., 2017). The main focus of researchers remain to secure the customers
data. With the innovations and technological revolutions each and everything has become
dependent on cloud computing, IoT, big data machine learning. But unfortunately these
systems don’t provide solid Service Level Agreements (SLA) (Marjani et al., 2017). Although
there are number of ways to enforce data security like encryption, temporary identification,
etc. But decisions should be made based on moral factors that is what to adopt, when to adopt,
where to adopt, how to adopt and why to adopt? (Marjani et al., 2017, Tsai et al., 2015). There
are number of existing solutions which are not pertinent (Marjani et al., 2017). They provide
solutions which are applicable for static data sets. But the in our case, of big IoT data have
dynamic nature (Marjani et al., 2017). In precision agriculture we cannot use the complex and
sophisticated algorithms due to limited resources like memory, processing, power
communication capabilities etc. In precision agriculture and farming the IoT sensing devices
are most defenseless especially to environmental tempering like attacks by animals, thieves,
and alteration of physical address (Varga et al., 2017, Duan et al., 2014). The IoT system when
used in agriculture are mostly targeted by device capture attacks (Varga et al., 2017, Newell
et al., 2014), the higher layers are then targeted by Dos attacks (Elijah et al., 2018). The
farmers usually prefer the devices which are cheap but they are not supported by security and
privacy related features. Hence to make internet of things more acceptable for every kind of
farmer there is a necessity of strict security and endowment policies for precision agriculture
(Fakhruddin, 2020).
Absence of Scalability and configuration: IoT system includes billions of devices which are
developed and deployed across the globe over different locations. Existing systems and
protocols should have the ability of support this large number of IoT nodes e.g. Sigfox gateway,
Lora and Ingenu can support up to 106 ,104 and 104 nodes respectively (Elijah et al., 2018).
This requires proper management system for each device and unique identification numbers.
Reliability: IoT sensing devices are mostly distributed in outdoor surrounding to sense the
environmental conditions. This exposes IoT system to harsh environmental conditions which
in turn destroys the sensors, gateways, etc. and hence creates the communication failures
(Elijah et al., 2018). Hence physical safety of these devices should be ensured so that big IoT
data precision system for agriculture will be properly developed.
Selection of technology: Number of challenges arises while developing a smart precision
agriculture system using IoT and big data. It is due to the fact that all the devices which exist
at the perception layer are directly exposed to harsh environmental conditions such as high or
low temperature, extreme humidity, storm, rain and many other possible dangers which
destroy electronic circuits and devices (Farooq et al., 2019). In precision agriculture to
maintain the system working in open fields is a challenging task. It is the responsibility of
researchers to address these issues.
Optimization of scarce resources: To determine the requirement of resources so that profit
margins can be known in advance but due to the varying nature of farm sizes and different
types of sensor requirement this is a challenging task (Elijah et al., 2018). In order to know the
requirement of resources precisely there is a need to develop complex algorithm and
mathematical models. Again the resources are limited in IoT system, e.g. Precision agriculture
advantages have already been discussed. However, the need of powerful gateways, smart
sensors, data centers and other IoT related gadgets can lead to a higher consumption of
energy (Fakhruddin, 2020). Hence it is very important to determine optimal resource
allocation while maximizing agricultural products and profits and minimizing the cost.
4.2.4. Challenges based on data analytics
Todays world is running every business and organization to get more profit using the data
analytics. Same is the case with the smart agriculture. To deal wit the data, researchers face
number of challenges which are mentioned under data analytics challenge category.
Integration: It refers to combine the data coming from different sources of different formats into
a uniform view (Ahamed et al., 2014). Good information is achieved from good data which is
achieved through proper integration (Liu and Zhang, 2014). But it is a challenging task to
integrate data from diverse sources (Ma’ayan et al., 2014). To adjust the structure in semi-
structured and unstructured data is again an issue to be addressed (Chen et al.,
2015, Savaglio et al., 2019). In Savaglio et al. (2019) novel data mining methodologies have
been investigated for IoT scenario with respect edge computing using two types of K-Means
clustering algorithm; distributed and centralized. Performance analysis helps to choose the
best among these two versions based on their computation, energy consumption and
communication requirements. There is a need to develop solutions for integrating data from
multiple sources efficiently.
Knowledge Mining: It refers to generate efficient and best fitting descriptive and predictive
solutions for big IoT data which in turn can be generalized for new data items (Marjani et al.,
2017, Mukhopadhyay et al., 2013). Information extraction and data exploration are the
challenges which were evolved due to big data and cloud computing platforms (Marjani et al.,
2017, Hu et al., 2012). Due to nature of big IoT data every task related to data mining becomes
a challenge like heterogeneous communication, integration, exploration and extraction
process, cleansing, reduction, transmission (Lerdsuwan and Phunchongharn, 2017, Khattab
et al., 2016, Liu et al., 2007, Ahmed et al., 2016). The researchers are trying to find solutions
to these issues. They have brought in sequential programming models, parallel
programming models and proposed diverse algorithms to deal with big IoT data. Big data is
increasing making precision agriculture data driven; this will become helpful only when we
make sense of it using technology.
Visualization: It is an important part in IoT data especially while handling IoT related systems
in which data is produced enormously (Marjani et al., 2017). Again it is a very challenging task
while dealing with big IoT data due to the structured, semi-structured and unstructured nature
of it. So we did not get good and clear results. Cloud computing platforms enriched with GUI
facilities help us to find good insights (Wang et al., 2015). Dimensionality reduction techniques
are very helpful to deal with the big IoT data (Azar and Hassanien, 2015). Again if we try to
use parallelization techniques to deal with small manageable tasks is again a challenge in
precision agriculture big IoT data (Childs et al., 2013).

5. Conclusion and future work

The arable land is reducing slowly due to constructions (houses, dams, factories, industries,
etc.), natural shrinking of arable land (floods, earthquakes, landslides etc.). Hence, the
emphasis on superior and proficient crop disease approaches is prerequisite so that
increasing food demand of the growing population can be fulfilled. The paper has given an
extensive literature survey on the precision agriculture using IoT data analytics and machine
learning. Furthermore, the applications of these technologies in precision agriculture are
highlighted. Moreover the paper tried to address one of the problems faced in apple crop in
regions like Kashmir. Since the apple growers are still using the classical approaches of
disease prediction without any technological intervention like IoT/ WSN. One of the major
advantages of the proposed approach will be real time measures against any possible disease
like scab if predicted precisely and in a timely manner. Results from a case study on adoption
of apple scab disease prediction in the orchards of Kashmir has shown improvement in terms
of dissemination of information regarding the factors/parameters responsible for the disease
aforementioned. Finally, the paper highlighted the issues faced while developing and
deploying such applications in real time, for this a local survey was conducted to know what
are the views of people regarding precision agriculture. Although many challenges are ahead
before implementing the proposed framework at large scale. These include initial cost of
implementation, deployment, training, weather conditions and other parameters. But once the
aforementioned limitations are overwhelmed, the profits will turn into visible and workable
form. The work shall be extended in future to include more parameters influencing the crop.
The paper generalize that precision Aggie is stand on four stakes; Appropriate source,
Appropriate place, Appropriate quantity and Appropriate time.
References
1. Abdmeziem et al., 2016
M.R. Abdmeziem, D. Tandjaoui, I. Romdhani
Architecting the internet of things: state of the art
Robots and Sensor Clouds, Springer (2016), pp. 55-75
View at publisherCrossrefView in ScopusGoogle Scholar
2. Aggelopoulou et al., 2011
A. Aggelopoulou, D. Bochtis, S. Fountas, K.C. Swain, T. Gemtos, G. Nanos
Yield prediction in apple orchards based on image processing
Precision Agric., 12 (3) (2011), pp. 448-456
View at publisherCrossrefView in ScopusGoogle Scholar
3. Ahamed et al., 2014
B.B. Ahamed, T. Ramkumar, S. Hariharan
Data integration progression in large data source using mapping affinity
2014 7th International Conference on Advanced Software Engineering and Its
Applications IEEE (2014), pp. 16-21
View at publisherCrossrefView in ScopusGoogle Scholar
4. Ahmed et al., 2016
N. Ahmed, H. Rahman, M.I. Hussain
A comparison of 802.11 ah and 802.15. 4 for iot
Ict Express, 2 (3) (2016), pp. 100-102
View PDFView articleView in ScopusGoogle Scholar
5. Ahmed et al., 2018
N. Ahmed, D. De, I. Hussain
Internet of things (iot) for smart precision agriculture and farming in rural areas
IEEE IoT J., 5 (6) (2018), pp. 4890-4899
View at publisherCrossrefView in ScopusGoogle Scholar