COTTON- PM- BACK

https://www.sciencedirect.com/science/article/abs/pii/B9780323852142000112#

Chapter 2 - A systematic review of artificial intelligence in

agriculture
Parvinder Singh, Amandeep Kaur

Department of Computer Science and Technology, Central University of

Punjab, Bathinda, Punjab, India

Deep Learning for Sustainable Agriculture: Cognitive Data Science in Sustainable

Computing. 2022, Pages 57-80
https://doi.org/10.1016/B978-0-323-85214-2.00011-2Get rights and content

Parvinder Singh, Amandeep Kaur,

Cite as- Chapter 2 - A systematic review of artificial intelligence in agriculture,

Editor(s): Ramesh Chandra Poonia, Vijander Singh, Soumya Ranjan Nayak,
In Cognitive Data Science in Sustainable Computing,
Deep Learning for Sustainable Agriculture,
Academic Press, 2022, Pages 57-80,

Abstract
The current world population is 7.8 billion and is projected to reach 9.8 billion by
2050. The limited land area and strong need to produce more crop to feed the ever-
increasing population is a major challenge today, especially for developing countries.
The strong need to produce more crop from lesser land has led to several
challenges in the field of agriculture. Reduction in agriculture yield due to climate
change and global warming due to farming has become a vicious circle. Excessive
use of chemicals in farms to increase soil fertility and reduce weeds and pests have
adversely affected the environment and the human health. There is limited
availability of natural resources like phosphorous and energy required in agriculture.
Water scarcity and increase in plant diseases are other major concerns.

Artificial intelligence (AI) has emerged as a promising technology in digital

agriculture. Digital agriculture relates to using digital technologies for collecting,
storing, and further analyzing the electronic agricultural data for better reasoning and
decision-making using AI techniques. Precision agriculture is one such technique
that monitors soil moisture and composition, temperature, and humidity and
determines optimized fertilizer and water requirements for a specific crop and
different areas of a farm. Then there are computer vision and machine learning
techniques to detect diseases and deficiencies in plants, recognizing weeds that
helps in spraying only those parts of land where the plants are disease-infected or
where weeds are present instead of the whole field. Utilization of AI in agriculture is
helping in developing agricultural methods capable of increasing crop yield and
reducing the previously stated challenges.
With merits of using AI, there are certain issues. The first major issue in using the AI
techniques is the need for high computational power that, again, leads to global
warming. Also, in developing countries, the internet infrastructure needs to be
improved to use AI techniques effectively. Cost of using AI is high, and countries
need AI experts to use the techniques to full potential. The focus of this chapter is to
review how AI techniques are helping in increasing yield and overcoming limitations,
like global warming, excessive use of fertilizers, limited availability of natural
resources, plant disease, and water scarcity. The chapter concludes by discussing
the issues and challenges in using AI, especially as it related to agriculture.

https://www.atlantis-press.com/proceedings/icamida-22/125986347
Advances in Computer Science Research

Proceedings of the International Conference on Applications of Machine Intelligence and

Data Analytics (ICAMIDA 2022)

Review of Machine Learning Model Applications in Precision

Agriculture
Authors

Patil Sagar Baburao1, *, R. B. Kulkarni2, Pramod A. Kharade3, Suchita S. Patil4

1
Department of Technology, Shivaji University, Kolhapur, India
2
Government College of Engineering, Karad, Maharashtra, India
3
Department of Computer Science and Engineering, BVCOE, Kolhapur, India
4
Shivaji University, Kolhapur, India
*
Corresponding author. Email: someone.sagar@gmail.com
DOI 10.2991/978-94-6463-136-4_81How to use a DOI?

Keywords

Machine learning; Crop yield; Disease detection; Support Vector

Machine; K-Means Clustering; ANN; CNN; Precision agriculture

Abstract

Over the past two decades, modern agriculture has made significant
advancements. The methods used in farming have changed from
conventional ways to digital technologies as a result of significant
technology improvement. Advances in machine learning and artificial
intelligence are being applied in this discipline to reevaluate farming
practices in order to meet the demands of an expanding population.
 Throughout the entire cycle of planting, growing, and harvesting,
machine learning is prevalent. It starts with the planting of a seed in the
ground, goes through soil preparation, seed breeding, crop health
monitoring, measuring water feed and concludes with the harvest being
picked up by robots by using computer vision techniques.

For crop selection, yield prediction, soil classification, weather

forecasting, irrigation system, fertilizer prescription, disease prediction,
and determining the minimal support price, machine learning models
are developed in the field of precision agriculture. In this article we
will cover the different categories of precision agriculture applications
and use of machine learning models in those different categories.
Various models in precision agriculture include Artificial Neural
Networks, Support Vector Machines (SVMs), Convolution Neural
Networks (CNN), Random Forest (RF), K-Nearest Neighbor (KNN),
K-Means Clustering. The ultimate solution to issues in agriculture rests
in the efficient application of Machine Learning (ML). ML can bring
about a paradigm change in nations like India where agriculture is the
main source of employment. Since most Indian rural areas have
adopted digitalization, ML and AI-related applications are gradually
emerging in this sector.

Copyright

© 2023 The Author(s)

https://www.academia.edu/43357425/
A_Review_on_Usage_and_Expected_Benefits_of_Artificial_Intelligence_in_Agricultu
re_Sector

A Review on Usage and Expected Benefits of Artificial

Intelligence inAgriculture Sector
Sanjiv Sharma and Jashandeep Singh
International Journal of Advanced Science and TechnologyVol. 29, No. 11s, (2020),
pp. 1078-1085
Chitkara Business School, Chitkara University, Punjab, India
Abstract
This paper is an attempt to synthesise the previous studies conducted by previous authors and the
major purpose is to explore the current trends of usage and expected benefits of Artificial
Intelligence (AI) inagriculture using systematic review. An extensive literature review has been carried by
identifying studiesconducted on usage and benefits of AI in the farming sector. The analysis reveals that is a
clear pattern of concentration of studies in three major areas such as crop and soil monitoring, predictive
agricultural analytics and supply chain efficiencies. Many such studies are more theoretical and are based
more on AI capabilities. No comprehensive research so far has been conducted to establish what could be the
best and impactful medium to deliver various observations derived from the application of AI tools.
Keywords:
Artificial Intelligence; AI; expert system; farm-base advisory

Introduction
The term ‘Artificial Intelligence’ (AI) has been a buzzword these days in several sectors, be it service,
manufacturing and agriculture. It needs to be noted that service and manufacturing sectors
have started toadopt AI practices and have been able to solve number of challenges in this competitive
environment.However, agriculture is sector still is in the evolving phase of using such AI practices. The
concerns in theagriculture sector are low income generation for the farmers, climatic
dependence, lack of awareness amongfarmers etc. But developed nations in Europe and USA have
been able to reap rich dividends due tointervention of AI in agriculture. Such AI based applications are used
for farm-based advisories regardingsprays, weather forecasting, usage of drones in the farms,
infrastructure for humidity and temperatureupdates to the farmers etc. Due to this the losses
of farmers have started to decline. Therefore, considering the aims of the government regarding
doubling the farmers’ income, the losses of the farmers have to beminimised using AI practices.
Considering this fact, the present study is an attempt to synthesise the previous studies conducted
by previous authors and the major purpose is to explore the current trends ofusage and
expected benefits of AI in agriculture using systematic review.

The industrial revolution of last two centuries which got driven by fossils and fuel poweredmechanization of
mining, manufacturing and its rapid expansion in other areas has brought societalchanges, mostly good in
nature yet with unintended consequences like urbanization and population growth,the challenge remained
how to feed and clothe the growing population unless agriculture production and productivity
could keep pace. To take on this challenge post World War - 2, a great emphasis has been put
for improving agriculture production. First in this line was “Green Revaluation” led by Norman Borlaug, in
last 50 years or so, world population got doubled and with the technology advances and
agronomicadvances, the production of cereal like wheat, rice and maize got tripled with a little increase in
cultivatedarea (Danny Llewellyn, 2018). In last 20 years have also seen introduction of
genetically modified seedswhich has reduced the usage of chemical pesticides and herbicides in key crop
which has helped in crop efficiency beside helping the sustainability. Technology and agriculture
research keep evolving ,same timet he world population keep increasing at faster rate than agriculture
production , pace is still on and it is predicated that by 2050 , the world population will be
around 9.8 billion , 30% more than today’s level andall this increase will come from
developing countries like India (foodtank, 2014).

The income level of these countries will keep rising hence challenge will be to feed larger,
urban and affluent mouths where thef ood habits, lifestyle and environmental challenges will be much
intense than now. Mounting challenge isthat in rapid era of urbanization, climate challenges and
industrialization how to feed the world of 2050.Do we need and green revolution? Answer is certainly Yes,
but this revolution should be smarter one,must provide scope, for the application of effective and efficient
agronomy solution, judicious use of inputs, more informed agriculture decisions, quantum jump in
productivity, more crop per drop, more producefrom field and a sustainable income to grower
Benefits of AI in Agriculture

Over the last decade, new learning, propositions, and new anticipated systems for
improvement inagriculture management ranging from the database to the decision-making
process has emerged (Bagchi2018). The research article by Bannerjee (2018) covers more
than fifty technological interventions for specific problems and solving them through expert systems. It
covers machine learning, fuzzy logic systemIoT in agriculture. AI usage and its impact in nine crops across six
countries have been analysed and it has been inferred that AI intervention in each crop
was effective in the range of 70-90 percent, however, theremaining 10 percent is observed as
a limitation. Naresh, Singh, (2014) revealed that the introduction of technology pays. The tec
hnologyadvancements in balanced fertiliser usage, chemical usage, sowing of high yielding
varieties andmechanisation of various agriculture operations including irrigation have
resulted in a significantenhancement in agriculture production.Singh and Chancellor

(1974) found that the tractor and tube well led mechanical intervention hasincreased the
wheat yield in Punjab, Haryana and Western UP significantly. The authors revealed
that thattractor owned farms recorded high productivity for paddy, sugarcane, and groundnut, from 4.1
percent to28.3 percent, 13.15 to 34.2 percent and 9.8 to 54.8 percent, respectively. This was
the impact of the firstwave of technology introduction which brought high productivity and efficiency in
farm management.According to Maher (2018), due to the increased usage of AI in the farming
sector, the AI usage inthe agriculture sector in the United States of America (USA) has shown significant
growth and is expectedto touch US$1.1 billion by 2025 from US$ 240 million in the year 2017. The AI
application is estimated toconnect 70 million agriculturists so that coordinated farm services will be gained by
2020 and increase theearnings of the farmers by US$ 9 billion. It will also help to resolve the issues related to
food security,animal husbandry, crop management, and soil maintenance.Artificial
Intelligence is powering agriculture by interpreting ,acquiring and reacting to the variousdata
it gets from agriculture field , the accurate solution information for each crop agronomy situation can be
delivered to farmers to take informed decision, 175 farmers of Andhra Pradesh attached to a
project
run by Microsoft India got an information support for efficient land peroration, sowing and n
utrientmanagement. Impact of this AI intervention, farmer got 30% more yield as comparison
to previous yearsFrom the above discussion, AI has been widely used practice in developed nations and
farmershave also gained dividends from this intervention by AI in agriculture, the scope and spread of
technologyis required for the developing and underdeveloped countries too. Therefore, it
would be an interesting factto review studies on several benefits and impact associated with AI. In the
next section, Usage of AI inagriculture are discussed.

Usage of Artificial Intelligence in Agriculture:

The significant growth in the farming sector in developed nations is attributed to the adoption ofAI based
expert systems. This means that there are several benefits associated with AI based systems
inagriculture. Therefore, there is need to understand the various benefits of such expert systems based on AIin
the agriculture by reviewing the findings of previous studies.Jairath and Yadav (2012) have
highlighted that taking informed decisions with the help ofInformation and Communication
Technology (ICT) is helping farmers across the value chain, promotingefficient production and
commodity trade. Mahindru (2019) examined the impact of AI and analysed asurvey conducted by NITI
Aayog, which described that AI machines provide relevant information to thefarmers about
best sowing practices, quality of soils, usage of pesticides, etc. As a result, the penetration
ofAI in agriculture is highly beneficial for farmers. For instance, the use of deep learning
techniques in Indianagriculture practices helps the farmers to make accurate crop yield
assessments and predict the overall production with more precision.
https://www.sciencedirect.com/science/article/abs/pii/
S0013935123024052

Breaking boundaries: Artificial intelligence for pesticide detection

and eco-friendly degradation
Diyasha Banerjee a, Satadal Adhikary b, Suchandra Bhattacharya c, Aritra Chakraborty a, Sohini Dutta a
, Sovona Chatterjee a, Abhratanu Ganguly a, Sayantani Nanda a, Prem Rajak a

Abstract
Pesticides are extensively used agrochemicals across the world to control
pest populations. However, irrational application of pesticides leads to
contamination of various components of the environment, like air, soil,
water, and vegetation, all of which build up significant levels of pesticide
residues. Further, these environmental contaminants fuel objectionable
human toxicity and impose a greater risk to the ecosystem. Therefore,
search of methodologies having potential to detect and degrade
pesticides in different environmental media is currently receiving
profound global attention. Beyond the conventional approaches, Artificial
Intelligence (AI) coupled with machine learning and artificial neural
networks are rapidly growing branches of science that enable quick data
analysis and precise detection of pesticides in various environmental
components. Interestingly, nanoparticle (NP)-mediated detection and
degradation of pesticides could be linked to AI algorithms to achieve
superior performance. NP-based sensors stand out for their operational
simplicity as well as their high sensitivity and low detection limits when
compared to conventional, time-consuming spectrophotometric assays.
NPs coated with fluorophores or conjugated with antibody or enzyme-
anchored sensors can be used through Surface-Enhanced
Raman Spectrometry, fluorescence,
or chemiluminescence methodologies for selective and more precise
detection of pesticides. Moreover, NPs assist in the photocatalytic
breakdown of various organic and inorganic pesticides. Here, AI models
are ideal means to identify, classify, characterize, and even predict the
data of pesticides obtained through NP sensors. The present study aims
to discuss the environmental contamination and negative impacts of
pesticides on the ecosystem. The article also elaborates the AI and NP-
assisted approaches for detecting and degrading a wide range
of pesticide residues in 0various environmental and agrecultural sources
including fruits and vegetables. Finally, the prevailing limitations and
future goals of AI-NP-assisted techniques have also been dissected

Diyasha Banerjee et al, 2024.

Breaking boundaries: Artificial intelligence for pesticide detection
and eco-friendly degradation, Environmental Research,
Volume 241, no. 117601,
Sensors (Basel). 2023 Apr; 23(8): 4127.
Published online 2023 Apr 20. doi: 10.3390/s23084127

A Review of Successes and Impeding Challenges of IoT-Based

Insect Pest Detection Systems for Estimating Agroecosystem
Health and Productivity of Cotton
Denis O. Kiobia,1,* Canicius J. Mwitta,1 Kadeghe G. Fue,2 Jason M. Schmidt,3 David
G. Riley,3 and Glen C. Rains1,3
Dinko Oletic, Academic Editor and Nahina Islam, Academic Editor
1
College of Engineering, University of Georgia, Tifton, GA 31793, USA
2
Department of Agricultural Engineering, School of Engineering Science and
Technology, Sokoine University of Agriculture, Morogoro P.O. Box 3003, Tanzania
3
Department of Entomology, University of Georgia, Tifton, GA 31793, USA
*
Correspondence: ude.agu@aiboik.sined or zt.ca.aus@sinedneglo; Tel.: +1-229-339-
7059 or +255-758384072

Abstract

Using artificial intelligence (AI) and the IoT (Internet of Things) is a primary focus of
applied engineering research to improve agricultural efficiency. This review paper
summarizes the engagement of artificial intelligence models and IoT techniques in
detecting, classifying, and counting cotton insect pests and corresponding beneficial
insects. The effectiveness and limitations of AI and IoT techniques in various cotton
agricultural settings were comprehensively reviewed. This review indicates that insects
can be detected with an accuracy of between 70 and 98% using camera/microphone
sensors and enhanced deep learning algorithms. However, despite the numerous pests
and beneficial insects, only a few species were targeted for detection and classification
by AI and IoT systems. Not surprisingly, due to the challenges of identifying immature
and predatory insects, few studies have designed systems to detect and characterize
them. The location of the insects, sufficient data size, concentrated insects on the image,
and similarity in species appearance are major obstacles when implementing AI.
Similarly, IoT is constrained by a lack of effective field distance between sensors when
targeting insects according to their estimated population size. Based on this study, the
number of pest species monitored by AI and IoT technologies should be increased while
improving the system’s detection accuracy.

Keywords: cotton pests, pest detection, pest control, machine learning, computer
vision, precision agriculture
 Go to:

1. Introduction

In recent years, the most challenging initiative for pest management decisions has been
assisting farmers in automated artificial intelligence sensor-based technologies and
Internet of Things (IoT) applications. A lack of artificial intelligent sensor-based
technologies and IoT leads to farmers attempting to scout the pests themselves,
enlisting the help of extension specialists and crop consultants to reduce the losses
associated with insect pest crop damage. The high crop loss rate (approx. 40%) due to
pests may be a consequence of scouters’ inability to detect insect pest buildup in both
production fields and nearby crops in real time [1,2,3,4]. There are many pests in the
global cotton ecosystem (approx. 1000 species) [1], and nearly 125 per country [2].
When scouting such numerous field pests without sensors, some species may be
practically impossible to detect with the naked eye [5]. The drawbacks of scouting
without artificial intelligent sensor-based technologies and IoT applications include
pesticide reapplication costs, pesticide resistance, health risks, and environmental-
ecosystem pollution [6,7]. In addition, manual scouting may lead to a lack of focus on
using selective pesticides and maintaining beneficial arthropods (predators and
parasitoids) as biological control agents [8,9,10].

Therefore, this review paper focuses on the perspectives and opinions of researchers on
the technical success and drawbacks of artificial intelligence and the Internet of Things
when detecting, classifying, and counting cotton insect pests. Cotton growers have
urged researchers to design Artificial Intelligence (AI) sensor-based systems and
communication technologies to identify, classify, and monitor pests [9,11]. However,
while many computing technologies have emerged, a review of the existing approaches’
current status, limitations, and pitfalls is needed to optimize future systems. For
example, the use of AI technologies, such as the K-nearest neighbors (KNN), logistic
regression, decision tree, support vector machine (SVM), and deep convolutional neural
network (CNN) models may accurately and precisely help farmers to detect, classify
insect pests and suggest the appropriate pesticides [12]. Such AI techniques are mostly
recommended as certain insect pests are tiny and hard to detect [13]. Additionally, AI
techniques are essential because of the potential to identify pests in images using low-
cost RGB cameras [14]. However, few review studies reflect the perspectives and
opinions of current research when designing/deploying AI approaches to detect,
identify, and monitor insect pests, specifically in cotton fields.
Go to:

2. Scope of the Review

This study was mainly confined to detecting and classifying insect pests in cotton
settings. The articles that suggested predicting pest existence without detecting them
were excluded from this study. Moreover, articles were not included except if detection
involved an AI approach, as this study intended to understand the potential of AI, its
associated challenges, and its success in the cotton environment. The existing literature
was searched using the following keywords: “cotton pests”, “pest features”, “Artificial
intelligence model in cotton pest detection”, “machine learning”, “IoT in the cotton
field”, “cotton insect classification”, “cotton insect identification methods”, “cotton pest
counting systems”, and “cotton pest image processing.” The literature associated with
the keywords was identified in the database and indexed sources, including Google
Scholar, Science Direct, connected papers, UGA Libraries Multi-Search, Web of Science,
and Scopus. The study also used Google’s search engine and YouTube to collect
information on the advertisement of related cotton pest detection systems, as the
literature covering the intended topic was limited. The evaluation of the retrieved paper
included the detected insect, data acquisition, utilized device, tested object detection
and processing algorithm/AI models, performed activities (detection, classification, and
counting) challenges, and system performance. The paper’s organization includes the
following sections: Section 3 begins by understanding the common pest recognition
features. Section 4 provides the commonly detected cotton pest species, approach, and
performance indicators during the implementation of detection or classification. Section
5 details the AI models currently tested for detecting cotton pests and their
corresponding successes and challenges. Section 6 discusses manufactured intelligent
systems/devices for identification, classification, and counting after detection and
classification. This section also describes typical sensors and traps involved in
monitoring cotton pests in the field. Section 7 shows AI in beneficial insects (predators
and parasitoids) for controlling cotton pests or assisting in flower pollination. Section
8 discusses the challenges of implementing IoT devices. Section 9 reveals the urgent
need for IoT applications in managing major cotton pests, and Section 10 provides
recommendations for further research.
Go to:

3. Pest Recognition Features

Obtaining the proper pest features is one of the essential prerequisites for effective pest
recognition. The most prevalent strategy has been grouping the pixels in images of
similar characteristics (image segmentation). The color distribution, morphology,
texture (entropy properties), and local characteristics are some of the features that may
be retrieved during the image segmentation process [15,16]. Color distribution is
crucial during insect classification since different species of insects and their body parts
come in multiple colors. In comparing and understanding the color of a specific insect
pest, color indexing is often used to compare the color of retrieved images with that of
the query [17]. The techniques to extract color features include a color histogram [14],
color moments [18], and a color correlogram [19]. According to [17], the histogram
technique has been a potential image signature due to its good accuracy and sensitivity
regarding pest position and orientation dynamics. However, the color histogram
technique has a limited ability to differentiate the spatial relationship between color
patches. In capturing the spatial relationship of color patches and the entire color
distribution, color correlograms and color coherence vectors have been proposed,
primarily when used together. The utilization of color moments has been effectively
employed in content-based image extraction systems and has been reported to be far
more reliable in defining color distributions than the histogram technique [20].
The morphology and local features include contour length, boundary diameter, area,
curvature, and perimeter. In contrast, local features involve the form factor, roundness,
aspect ratio, compactness, and extension [21,22], as shown in Table 1. To be considered
a key feature, it must have a low degree of association with other categories, have a
more considerable relevance when seen by humans, and represent a significant
variation compared to other insect categories. On the other hand, the analysis of
textural properties involves contrast, correlation, entropy, energy, and homogeneity
properties, especially utilizing the image gray-level matrix [23].

Table 1

Essential techniques for the evaluation of morphological variations between insect

species during feature classifications.

S/No. Morphological Insect Feature Formula

1 Form Factor =(4 × π × Area)/(Perimeter)2
2 Roundness =(4 × Area)/(π × Max Diameter2)
3 Aspect ratio =(Max Diameter)/(Mean Diameter)
=(Sqrt ((4/π) × Area)/Max
4 Compactness
Diameter)
5 Extent =Net area/Bounding rectangle
Open in a separate window

Source [21].

When distinguishing one insect pest from another, it is critical to choose the crucial
criteria carefully to avoid using features that are not necessary, which can lead to
complicated algorithms.
Go to:

4. Detected Cotton Pest Species, Approach, and Performance Indicator

4.1. AI Performance Indicators

The performance of AI models such as Convolutional Neural Networks (CNN), Support

Vector Machines (SVM), K-Nearest Neighbors (KNN), Artificial Neural Networks (ANN),
RNN (Recurrent Neural Networks), DBN (Deep Belief Networks), and DBM (Deep
Boltzmann Machine) is dependent on the sensor’s data clarity. The data collected by
sensors and then processed using artificial intelligence models must be accurate to
provide meaningful information [24]; the best accuracy depends on good sensor
calibration before data collection [25]. As demonstrated in Table 2, the performance of
artificial intelligence models is commonly tested for accuracy (%), precision, recall
(sensitivity), and F-scores [12,21].
Table 2

Common techniques to measure the performance of artificial intelligent models during

image processing.

No
Analysis Term Formula Description
.
=[(TP + TN)/(TP + FP + Estimates the percentage of correct
1 Accuracy (%)
FN + TN)] × 100 predictions made by a model
Indicates the quality of a positive
2 Precision =[TP/(TP + FP)]
prediction made by the model
Recall Evaluates how accurately the model is
3 =[TP/(TP + FN)]
(sensitivity) capable of identifying the relevant data
Calculates the model’s overall accuracy
=2/[(Recall)−1 +
4 F1-score by combining the precision and recall
(Precision)−1]
metrics in a twofold ratio.
Mean Average =((∑nq=1AP(q))/Q) × Shows the Average Precision metric
5
Precision 100% obtained from Precision and Recall
Open in a separate window

True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), Total
pest categories (Q), and the average accuracy rate of results detected in each pest
category AP (q) [11,26].

4.2. Detected Cotton Pest Species

The names of the cotton pests were identified in various research using AI image
processing techniques, as shown in Table 3. We found that some pests, such as Cotton
Whiteflies, Spiders, Pink bollworms, and American bollworms, were proven to be
detected using AI models in multiple studies. Other pest species were only examined
once in a single study. The choice to include a particular pest in this study may be based
on the availability of image data to meet the model’s requirements, which are chosen for
that particular study.

Table 3

Summary of recently detected cotton pests using AI.

Detected Insect Pests Reference

Boll weevil, Cotton aphid, Cotton bollworm (larva), Cotton bollworm
(adult), tobacco budworm (larva), Tobacco budworm S (adult),
Soybean looper, Fall armyworm (larva), Fall armyworm (adult), [1]
Cotton leafworm, Cotton whitefly, Cotton bug, Pink bollworm,
southern armyworm, and red spider mite
Cotton aphids, Flea beetles, Flax budworms, and Red spider mites [11]
Detected Insect Pests Reference
Mexican cotton boll weevil, Fall armyworm, Cotton bollworm,
Cotton aphid, Cotton whitefly, Green stink bug, Neotropical brown [14]
stink bug, Soybean looper
Assassin Bug, Three-Corned Alfalfa Hopper, and Convergent lady
[21]
beetle
Red spider mites and Leaf miner [27]
Pink and American bollworms [28,29]
American bollworm, Ash weevil, Blossom thrips, Brown cotton
moth, Brown soft scale, Brown-spotted locust, Cotton aphid, Cotton
leaf roller, Cotton leafhopper, Cotton looper, Cotton stem weevil,
Cotton whitefly, Cream drab, Cutworm, Darth maul moth imago,
Darth maul moth, Desert locust, Dusky cotton bug, Giant red bug,
Golden twin spot tomato looper, Green stink bug, Grey mealybug,
Hermolaus, Latania scale, Madeira mealybug, Mango mealybug,
[29]
Megapulvinaria, Cotton stainers, Menida, Menida-versicolor, White-
spotted flea beetle, Myllocerus-subfasciatus, Sri Lankan weevil,
Painted bug, Pink bollworm, Brown-winged green
bug, Poppiocapsidea, Red-banded shield bug, Red cotton bug, Red
hairy caterpillar, Solenopsis mealybug, Spherical mealybug, Spotted
bollworm imago, Spotted bollworm, Tobacco caterpillar Tomentosa,
Transverse moth, Tussock caterpillar, and Yellow cotton scale
Cotton whitefly [1,14,23,29,30,31,32]
Open in a separate window
Go to:

5. Tested AI Models for Pest Detection

The studies of artificial intelligence (AI) for image-based cotton pest detection are
shown in Figure 1. Although it is estimated that there are more than 1000 pests in the
cotton environment, very few have been identified using AI (Figure 1a). However, using
AI to identify cotton pests is promising based on best-reported accuracy in several
studies (Figure 1b). According to our research, the range (71.7–98.9%) was the most
frequently reported high detection accuracy across numerous investigations. Figure 1c
displays a comparison of widely mentioned accuracy classification algorithms/models.
Multiple studies investigated Faster R-CNN extensively and found it reasonably
accurate. Although the studies were independent, typical CNN, Few-Shot Learning, and
Single Shot Detector improved ResNet34 (ResNet34∗), SegNet, ANN, and MATLAB
classifiers, which demonstrated a reasonable detection accuracy in cotton pest image
classification. Furthermore, it was found that these effective image classification models
were reported primarily by comparing multiple models (Figure 1d). Most of the
research used upgraded CNN (Few-Shot Learning, Single Shot Detector, ResNet34, and
SegNet) to attain maximum accuracy.
Figure 1
Summary of AI image processing algorithms used for detecting cotton pests. (a) The
number of cotton pest species included in the study. (b) The highest accuracy reported
in the study. (c) Comparison of accuracy of image classification models reported in
cotton pest detection studies, and (d) = the number of classification models involved in
an individual study. The references A, B, C, D, E, F, G, H, I, J, and K represent references
[1,11,14,21,23,27,28,29,30,31,32], respectively. The (*) indicates improved ResNet34.

For example, using CNN models, fifteen (15) cotton insect pests were classified utilizing
100 original RGB color images and 620 augmented images for each insect [1]. The study
controlled the augmentation process by rotating the original images in 10-degree
increments and performing random sliding, stretching, and zooming up to 1.5×. The
ability of the four CNN models: AlexNet, ResNet34, ResNet50, and ResNet34∗, were
tested. ResNet34∗ was a significant attempt and contribution made by the authors to
improve ResNet34 and, specifically, to attain a higher accuracy in cotton pest
classifications. The CNN models were compared to the LBP-SVM model, which was
developed using Linear Binary Features (LBP) and a Support Vector Machine for the
baseline comparison. Although all of the CNN models in this study had better accuracy,
the author claimed that the CNN models with residual architectures (AlexNet, ResNet34,
and ResNet50) had more accuracy than the typical CNN model (AlexNet). Furthermore,
the improved ResNet34 (ResNet34∗) demonstrated greater accuracy than the other
evaluated models. The achievement of improving the ResNet34 models indicated the
potential that the original CNN models might be modified for improved identification of
cotton insect pests. One of the most significant challenges identified in this study was
the models’ inability to classify images of insect pests with similar appearances, such as
Heliothis virences (adult) and Spodoptera frugiperda (adult). Such a barrier may
emphasize the researchers’ need to collect more data during detection rather than
relying solely on images, as revealed in [33,34,35,36,37].

The CNN technique, called the “few-shot learning approach”, was developed to detect
and learn cotton pests using a few images instead of many images as an input data set
[24]. The goal of upgrading common CNN deep learning structures was to minimize the
image processing time and effort spent gathering a big data set and applying high-
powered or extensive field-hardware computing resources such as GPUs and TPUs
during processing. Additionally, the authors wanted to avoid existing image pre-
processing procedures such as target cropping, grey transformation, and resizing [38].
The suggested AI algorithm was also useful since, unlike GPUs, TPUs, and servers, it is
compatible with low-cost embedded hardware that may be used in the field. After
comparison, the authors found that the few-shot model outperformed other standard
models such as GoogleNet, AlexNet, VGG-16, VGG-19, ResNet-50, and ResNet-101 when
classifying cotton pests. This indicated that when embedded in running devices, the
few-shot learning model may be considered one of the significant successes and
promising techniques in image processing. This is because, when addressing some of
the shortcomings of existing models, such as using a few training images, the need for a
robust model training machine/computer results in the need to reduce data collection
and model training time. Similar systems that swiftly compare the metric space between
characteristics of input images have been proposed in Siamese network relation
networks and prototype networks [39,40]. Despite an improved accuracy, the focus was
on improving total model performance through a parallelism technique, which supports
partitioning them into simple tasks and performing them simultaneously and
accurately.

A comparison was conducted between a Deep Convolutional Neural Network (DCNN),

Hierarchical Deep Convolutional Neural Network (HD-CNN), and Pixel-wise Semantic
Segmentation Network (SegNet) to detect five cotton pests [14]. The authors intended
to modify the SegNet and compare its performance with the typical (DCNN) and (HD-
CNN) as a baseline. The performance accuracy of the SegNet model in identifying and
classifying the five tested cotton pest insects was reported to overtake (DCNN) and (HD-
CNN). They used pixel-wise classes that resulted in lower errors than the other
approaches. Such a performance aligned with the need to merge the improved model
with a robot that could save the obtained data set images, detect pests, produce and
record coordinates using geographical navigation satellite systems (GNSS), and
undertake insect pest mapping. More information on the influence of the image
background, specifically during the separation of the image background and the
targeted insect, is still an issue to investigate further. Furthermore, the authors
expressed the need to choose the optimal number of automated convolutional layers
optimizing training pace and model performance.

An application that includes an Internet of Things (IoT) device and a faster region-based
convolutional neural network (Faster R-CNN) to classify insect pests through cloud
computing was reported in [11]. This application utilizes a deep-learning model to send
images of unknown crop pests to the cloud via an iOS or Android-based smartphone for
storage, identification, and classification. The Faster R-CNN was quicker than previous
image classification techniques such as the Single-Shot Multi-Box Detector (SSD)
MobileNet and backpropagation (BP) neural networks. The accuracy (98.9%) of this
suggested technique exceeded that of SSD MobileNet (86%) and the BP neural network
(50%). The following are some of the promising contributions of the proposed model:
detecting insect pests on complex visual backgrounds, recognizing pests in real-time,
and forecasting agricultural pest classes and locations using the RPN module (Region
Proposal Network).

Furthermore, the suggested mobile application and cloud computing were significant
since they added the capability of proposing appropriate pesticides after the pest was
classified. However, the number of pests examined (five) was insufficient. There is a
need for future research to study a similar system with an increased number of cotton
pests. Moreover, the effect of different mobile phone resolutions, varying illumination
as crop growth conditions, and seasonal weather may need to be sufficiently
investigated.

In a similar study, 120 images were used to compare the performance of Fast R-CNN
and YOLOv4 detection models [32]. The Faster-RCNN model had a better accuracy of
95.08% (F-1 Score: 0.96, recall: 98.69%) compared to a YOLOv4 model that indicated a
lower accuracy of 71.77% (F-1 score: 0.83, recall: 73.31%). However, the YOLOv4
model was faster than the Faster R-CNN.

Moreover, a CNN architecture was shown to identify and classify spider mites and leaf
miners by collecting 600 RGB-colored images of insect pests [27]. The authors utilized
the K-fold confirmation method to divide and upgrade the CNN model. The authors
accomplished the deep learning process by combining Keras, TensorFlow, and Jupyter
deep learning libraries. The developed model was 96.4% accurate for the tested classes
after 100 epochs compared to 50 or 150 epochs. However, compared to the study by
[1], which required 100 original images and 620 augmented images per insect category
to achieve 98.1% accuracy using ResNet34∗, the work in [27] utilized substantially
fewer images (600) to achieve 96.4% accuracy. This study also presented similar
challenges, including complexity in the insect’s features (shape), dynamics due to light
intensity, insect orientation, various backgrounds, and similarity in insect size.

In 2000 a Deep Neural Network was used to classify and differentiate insects in the
cotton ecosystem [21]. The insect pests were classified using the decision tree method.
The study involved eight (8) extraction features, including insects with and without
legs. One of the most critical findings of this study was that the model could not
accurately recognize all the insect pests. Four insect species were detected at 77.0, 85.7,
91.5, and 95.6% accuracy. According to the author, one of the reasons for poor
classification was the high light reflectance of some insect pests. Despite providing a
framework for cotton insect classification, the findings were not effectively applied or
expanded until the expansion of deep learning studies in 2006 due to limitations in the
computer system and neural network theory during that period [41,42]. The studies
involving basic Deep Neural Networks were limited.
Go to:

6. Intelligent Sensor Systems for Monitoring and Counting Cotton Pests

6.1. System Components of Remote Monitoring Devices

Remote monitoring platforms may include data collection, storage, analysis, and
information dissemination [43]. The information-gathering system can incorporate
several types of sensors, such as RGB, infrared, and hyperspectral cameras, to acquire
images of insect pests. The resolution should be considered when choosing a camera, as
a low resolution may negatively affect the image processing output [44]. Furthermore,
environmental data, such as light, rainfall, soil moisture content, and underground and
surface temperature sensors, may also be included in the information-gathering system
for immediate pest detection and the prediction of population changes [43]. After
collecting and handling data, data analysis was required to create valuable information
that could be shared with users to alert them to take pest management action. Such
information should be integrated with location information using GNNS to allocate the
field site under pest damage or pollination [9]. With artificial intelligence models, the
analysis section should be integrated with robust machine learning models that can
immediately correlate the sensor data and provide pest management recommendations.
For advanced pest tracking systems, the platform should provide multiple sensor data
and work in two ways: it should deliver field information to the user and allow the user
to query it at any moment, as illustrated in Figure 2.

Figure 2
The components of the insect pest monitoring platform. (a) System for information
collection, and (b) System for information analysis and display (adapted from [43]).

In cotton fields, the systems to detect flying insects in cotton fields named the “self-
cleaning trap” (Figure 3a) and “remote whitefly monitor” (Figure 3b) were shown in
[32] and [45], respectively. The “remote whitefly monitor” and “self-cleaning trap” use
sticky traps to fix the insects to the camera sensor’s position. However, in the “remote
whitefly monitor”, the attractive mechanism of pests to the camera is the sticky trap
(yellow, blue, or white color) itself, while in the “self-cleaning trap”, the pests are
attracted to the camera position using pheromone. In addition, the whiteflies monitor
device includes a function for counting pests and reporting online one to three times a
day. Both devices have the potential for further improvement, particularly in attracting
multiple pests and conducting specific insect classifications. These devices can also
potentially include models that reasonably forecast pest dynamics, including weather
conditions.

Figure 3
Remote monitoring devices for cotton pest identification. (a) Self-cleaning trap [45],
also shown at https://app.efos.si/trapview_help/html/Trapview_SCM.html (accessed
on 22 November 2022). Permission granted, and (b) Remote whitefly monitor [32].

6.2. Counting of Pests on Leaves

Capturing pest images directly from leaves may avoid the cost of purchasing insect
pests’ attractants, such as yellow sticky traps or pheromone agents. The technique for
identifying and estimating whiteflies’ population directly from the cotton leaf was
demonstrated [23]. This approach utilized MATLAB field-based machine learning
techniques. The data collection and analysis pipeline included image acquisition, the
conversion of RGB images to HSV (Hue, saturation, value), background removal, grey
color thresholding, and whitefly counting using a bounding box and region props
algorithm. The proposed technique for pest counting was both quick and cost-effective.
Additionally, the method was robust, indicating an accuracy rate of 98%. This approach
should also output the number of pixels, ratios, interior density, standard deviation,
skew, and kurtosis in the event of multiple species detection after the feature extraction
procedure [1]. However, image processing for whitefly recognition and counting in the
field is still loaded with difficulties, especially in dust, excess moisture, weather
conditions, and leaf veins that often indicate a light color similar to whiteflies. Moreover,
because whiteflies are smaller, detecting and analyzing them becomes more
challenging, mainly when individuals stick to each other [46]. In addition to insects’
small adhesion size, image processing becomes more complicated when the captured
cotton image has poor quality [47].

Another study included transferring machine vision models to a mobile device App for
agronomists to acquire images, analyze, and count Silverleaf whitefly nymphs on the
cotton leaf [30]. The method used segmentation and machine learning. Silverleaf
whitefly nymphs were recognized with up to 67% and 79% accuracy using
segmentation-based and Faster R-CNN -learning approaches, respectively. More work is
still being performed to make the device suitable for aphids and mites in cotton fields.
The suggested device development process was reported in [31]. Smartphones from
Apple, Sony, and Samsung were tested as part of the concept testing. The updates
indicated that the greatest F-scores of 71.7 to 75.8% were found in deep learning
models that used the iPhone alone or in conjunction with other smartphone models. All
deep learning models that included Sony and Samsung combined or separately without
integrating the iPhone model were reported with lower F-scores ranging from 44.0 to
55.6%. The model’s performance on the tested mobile smartphone missed the specified
F-score mark from 90 to 95%. The inability to achieve the necessary F-score threshold
was related to reduced image quality from using only one image sensor for training,
detecting, and assessing performance in a real-world cotton field scenario.

Similarly, the classification of pests on plant leaves based on a smartphone running iOS
13.5 through cloud computing techniques, the Internet of Things (IoT), and a faster
region-based convolutional neural network was demonstrated [11,30]. However, the
complexity of the resemblance between the plant’s leaf backgrounds and some pests
was still challenging. In encountering some of these problems, the utilization of high-
resolution digital devices is essential though it is expensive to purchase them. The
primary and minor axis lengths and eccentricity are widely employed to eliminate leaf
veins from images [48].

6.3. Counting Pests on Sticky Traps

The use of sticky traps is essential in minimizing detection challenges between

complicated backgrounds and the targeted insects. The sticky traps assist in sticking
and fixing the insects while waiting for the image [49]. The procedure for automatically
estimating the number of cotton pests, particularly whiteflies, using sticky traps is
broadly suggested in [23,32,50]. The method may include image acquisition, color space
conversion, background subtraction, thresholding operations, and morphological
tracking/labeling operations. A camera may scan the insects on the non-automatically
rotating sticky trap [50] as expressed (Figure 4a) or by automatically rotating the sticky
trap roll [32] as shown (Figure 4b–d). Additionally, insect feature extraction from sticky
tape may include training the model with labeled insect features using box boundaries
or a background removal approach. The technique for automatically rotating the yellow
stick roller in the morning and afternoon and shooting the trapped cotton pests was
also demonstrated [32]. The proposed system could detect pests on yellow sticky tape,
send images to a web server, count the pests, and send the insect count status to a web
server. This study demonstrated the possibility of employing the Faster-RCNN detection
and counting model, embedded computers, cameras, and sticky tape to identify and
count real-time cotton pests. This system could potentially submit daily pest monitoring
reports from multiple remote stations. Since the remote devices operate on solar energy
and low-cost batteries, the technology may be feasible for field operations, particularly
in developing countries. Automating rotating yellow tape may also be advantageous
owing to minimal maintenance, particularly in lowering the time required for the field
replacement of fixed stick traps. However, the training of the reported system relies on
detecting and counting only whiteflies. Improving such a system to detect and classify
multiple insect pests could significantly contribute to pest management. Moreover,
since such devices may operate autonomously in the field, integrating other
characteristics, such as micro-climates, may aid farmers in forecasting insect dynamics.

Figure 4
Photographing insect pests on the stationery and automated rotating sticky trap. (a)
Pest detected on a non-automatic rotating sticky trap card, (b) Insects detected on the
rotating sticky roller, (c) Sticky trap roll for an automatic rotating sticky trap, (d) Sticky
trap under the automated rotating system (adapted from [32,50].

6.4. Counting Pests on Paper

This approach collects the trapped insects onto a sheet of paper for imaging. The
trapping of insects involves attracting agents such as pheromones or light. Using a
smartphone camera, a system that takes images of insects trapped by pheromones was
studied [28]. The system processed the images and counted the pests using a Single
Shot Multi-Task Detector (SSD) supervised Neural Network algorithm. The technique
had the added benefit of indicating if the insect populations warranted spraying with an
insecticide. Due to agro-climatic zones, the study acknowledged and addressed the
issues of using low-resolution cameras and data diversity in cotton fields. This system,
however, was only taught to detect a few pests, including pink and cotton bollworms.
Because the system relies on human operation and the usage of mobile phones, human
error and differences in the versions of mobile phone cameras might cause variances in
the system’s performance. One strategy to avoid the complications of a smartphone
camera has been to kill the pest in the trap without crushing it before acquiring the
image. Such a practice may need a trained and committed farmer or consultant. In
addition, the optimal number of traps to capture pests per area was not explicitly
mentioned for a farmer to apply a pest management solution.
Go to:

7. Artificial Intelligence in Beneficial Insects

The perspective of detecting or identifying beneficial insects involves crop pollinators

(e.g., wild/honeybee and butterfly), predators (e.g., Chrysoperla externa Hagen, Eriopis
connexa (Germar), Podisus nigrispinus (Dallas) and Orius insidiosus (Say)) [8,51], or
parasitoids (e.g., Aphelinus gossypii (Timberlake), Bracon
vulgaris (Ashmead), Lysiphlebus testaceipes (Cresson), Telenomus podisi (Ashmead),
and Trichogramma pretiosum (Riley)) [51]. Detection and classification were conducted
based on images of morphological features [44,52] or insect-swarming activities
[53,54,55]. For example, when using bee morphological features with more than 9000
images under deep convolutional networks (CNN), the species and subspecies of wild
bees were automatically recognized with accuracy rates of 99.3% and 98.05%,
respectively [44]. Such accuracy findings were approximately similar to 98.5% reported
in [56] when classifying and counting the Hymenoptera parasitoids of Aphelinidae,
Braconidae, and Aphidiinae using Mask R-CNN. However, some classifiers did not
perform better on the morphological features of beneficial insects. For example, poor
accuracy levels (max. 65.15%) of bee’s morphological classifications were reported
using Naive Bayes, Linear Discriminant Analysis (LDA), Logistic, Support Vector
Machine (SVM), Multilayer Perceptron (MLP), KNN, and Decision Tree (C4.5) classifiers
[52]. The capacity of deep CNN algorithms to automatically extract and learn traced or
manipulated features straight from the data could be credited with increased accuracy
when utilizing them [57]. The downside of this approach is that it needs many images or
insect labels to achieve a satisfactory performance [43]. For example, over 9000 images
and over 15,000 insect labels were utilized in [44,56] to achieve over 90% higher
accuracy.

On the other hand, using acoustics is an alternative to the morphological feature

extraction technique. Deep learning and IoT-based approaches for monitoring and
categorizing bees based on the sound frequencies of their swarm activity were
demonstrated [53]. While the compressed audio (MP3) lost 10% accuracy, the
uncompressed audio showed a greater accuracy of almost 94%. By utilizing an
embedded computer, the study had the potential to pinpoint where beehives were
located. One of the challenges was delivering audio data continuously due to the
beehives’ remote location. Creating MP3 files solved this problem by taking full
advantage of the optimum cellular network bandwidth (16–64 kbps) for a few kilobits
per second rather than 256 kbps for the uncompressed audio file. A similar acoustic
machine-learning approach included additional features for forecasting and mapping
the pollinated crop field area [9]. The sound frequencies were grouped into flying,
fanning, hissing, and pipping at 250 Hz, 225–285 Hz, 3000 Hz, and 340–450 Hz,
respectively. The analogy detection filter removed the audio fragments before sending
them to the cloud for SVM processing. However, achieving an optimal number of
microphone sensors per m2 and the proper distance between the microphone sensor
and the flying insect remains challenging. For training and testing, the technique may
also necessitate collecting massive amounts of audio recording data on swarm activities
[54,55]. The option for obtaining suitable and sufficient sound recordings was to use
open-source data [53]. The same acoustic detection approach was developed for flying
insects such as bee pollinators [28]. The developed system included other field
variables such as light, rainfall, soil moisture content, ambient temperature, surface
temperature, and subsurface temperature. The system combined a raspberry pi
processing unit, a raspberry pi camera, a YOLO object detection model, and a data
server for monitoring flying insects. However, unless paired with a Support Vector
Machine (SVM), the authors found the YOLO detection model’s capabilities in classifying
flying insects limited. Unfortunately, studies that detect, classify, and quantify beneficial
insect species using deep learning techniques, especially predators, are still limited.
 Go to:

8. Challenges to the Implementation of IoT-Filled Devices

The challenges that remote devices for pest detection in the field may encounter include
high power consumption, network issues, inadequate security, service expiration,
physical hardware defects, software failure, and changes in ambient conditions, as
demonstrated in Table 4. The origin of a particular challenge on a device may change
occasionally and depend on the design of that device. If necessary, the suitable
operating IoT device installed on-site should be characterized by high performance with
minor faults. In the process of avoiding drawbacks for field devices, Table 4 also shows
the proposed solutions of different studies.

Table 4
Challenges of artificial intelligent field devices, sources, and proposed solutions.

Challenge Source Solution

Power  -  -
consumption High-power requirements of
networking devices, micro- Scheduling tasks, creating intelligent
controllers, and embedded software with the fewest
computers [58,59] computations possible, and making
idle mode.

 -

Low data rates may be achieved with

good energy economy when short-
range communication is used at
transmission distances of less than 20
m [59]

 -

Based on the installation site and

operation goal, IoT sensor nodes can
be set up as reduced-function gadgets
that only talk to full-function devices.

 -

Convey data to the control center

from intermediate nodes that receive
it from other IoT nodes [59]
Challenge Source Solution

 -

Solar power and other alternative

energy sources [59,60]

 -

Predicting power usage with energy

forecasting devices and models
 -
Limitation of software and poor
 -
Failure to execute processing power in case of
Comprehensive software testing
software rapid data processing over wide
before deploying
range of sensor data sources
[59,61]
 -
 -
Payment renewal of application
Service expiry Applications failing because of
services such as internet bundle and
fault expired or terminated cloud
app services should be configured to
services [62]
the automatic payment mode
 -
 -
Network breakdowns, packet
loss or corruption, congestion, or
The design should include fault
a problem with the destination
tolerance [61,62,64]
node [61,62]
Network faults
 -
 -
The design should include retrying
Inadequate Internet coverage
failed packages and real-time
causing connection issues that
reporting [65]
result in lost or incorrectly sent
data [63]
 -

Access control, authentication, and

 - authorization techniques [66]
Security Security for devices and data
privacy [61]  -

Make data encryption before

transmission
Physical faults of  -  -
hardware Issues with sensors, processors, The device should be calibrated and
memory, storage, and power undergo robustness testing before
Challenge Source Solution
supply [61] deployment [67]
 -
 -
Data cost Reduce the frequency of sending
High cost of transmitting images
images
 -
Changes in Extreme weather conditions  -
environmental such as rain, extremely hot or Add weather check sensors and
conditions low temperatures, wind, and modules [68,69]
other extreme situations
Open in a separate window

Go to:

9. Urgent Need for IoT towards Worldwide Major Cotton Pests

This study found that sensor-based artificial intelligence and IoT may need to devote
more effort to major pests when implementing pest management strategies. The
priorities may be determined by the worldwide economic losses posed by the individual
type of pest. For instance, whiteflies have been reported in Central America, North
America, South America, China, Central Africa, South Africa, South East Asia, Pakistan,
and India [70,71]. Similarly, cotton mealybugs have been reported in more than 26
countries in various ecological zones [72]. Boll weevils have been reported throughout
southern Texas to Argentina [73]. Several studies have indicated a tremendous increase
in serious destructive pests within the cotton ecosystem. For example, in the USA, the
cotton fleahopper was ranked ninth in early 1999 and fourth or sixth in 2007 in the last
decade [74]. The present study highlights the major pests and corresponding yield
losses reported in different studies (Figure 5). The magnitude of yield losses in cotton
yield may differ based on the local production practice, the type of pests, pest
population, the stage of the crop, and pest-supporting conditions in given ecological
zones of a specific region. For example, depending on the stage of the crop and the
infestation intensity in the field, fall armyworm larvae may cause yield losses of 25.8 to
100% [75,76]. Given the maximum losses likely attained by the most destructive pests,
cotton farmers may lose even more than 50% of their yield when attacked by such
major pests. In case of an attack by a major pest such as a boll weevil, farmers may
apply up to fifteen (15) insecticide sprays per season because of clear decisions or
guidance about the insect population threshold estimation [76,77]. Unfortunately,
managing major cotton pests is complex as most farmers and advisors use informal
knowledge to control the dynamics of such pests [78,79,80,81,82,83]. The recent shift
from deep fallowing/plowing to reduced or no-till farming practices has also been
linked to severe pest prevalence and attack consistency [84]. The severe pest attack was
associated with fewer farmers (approx. 8.33%) engaging in traps for the mass trapping
of pests [85]. Farmers have repeatedly been unable to control the severe pest attack
because of the pest’s ability to migrate from multiple crop hosts to cotton and a
temperature that favors pest reproduction [86]. Preventive seedling protection has
been recommended as one of the main approaches to reduce cotton infestations and
encounter the overuse of pesticides in the case of unknown pests. However, this
approach may adversely affect nontarget insects, including honey and other pollinators
[87,88]. Creating a database of pest images and linking it with artificial intelligent
models or sensor-based systems could easily classify multiple pests and recommend
insect-specific insecticides to reduce nontarget insects’ sublethal adverse effects.

Figure 5
Serious pests in cotton fields (i) and corresponding yield lost per pest category (ii). In
(i), a: mealybug, b: silverleaf whitefly, c: cotton leafhopper (adult), d: cotton aphid, e:
two-spotted spider mite (adult), f: cotton thrips, g: fall armyworm (larva), h: pink
bollwormi, i: fleahopper, and j: boll weevil (also shown
at https://www.insectimages.org/ (accessed on 11 December 2022)). In (ii), A: Boll
weevil [73], B: Cotton Thrips [78], C: Spider Mites [81,82], D: Fleahopper [74], E: Cotton
aphid [79,80], F: Leaf hopper [83,84,85,86], G: Cotton mealybug [72,89,90,91], H:
Whitefly [92,93,94], I: Pink bollworm [95,96], and J: Armyworm [75,76].
Go to:

10. Recommendations for Future Research

The studies investigated here focused on detecting cotton pests and beneficial insects
(predators and parasitoids) based on the images from the stems, undersides of leaves,
the root of the bloom, the external or internal part of the bloom, and roots are still
limited. Our review uncovers a scarcity of studies on detecting the insects that may
harm plants, which spend a portion of their life or growth cycles in the soil. In addition,
identifying pests from images with a complex background, concentrated problems, and
similar-looking species in one image is a major challenge when identifying cotton pests.
Based on this review, more studies should concentrate on possible attractants for such
pest species and utilize the sticky board to capture them for imaging in the sensor
region.

Attention should be given to some insects that may not have a long flight range, may not
fly at all, or, as mentioned, spend a portion of their life cycles underground.

Moreover, most artificial intelligence techniques have employed CNN models to detect
and classify cotton pests and beneficial insects. Nevertheless, using a typical CNN
classifier, on the other hand, may need a large quantity of training and validation data,
necessitating more time, more significant research expenditures, and compelling
operational hardware. To identify cotton pests with little data, training time, and low-
cost hardware, researchers should apply more efforts to understand the potential of
Small CNN architectures such as AlexNet, GoogLeNet, Inceptionv3, SqueezeNet, ResNet-
101, VGG16, ShuffleNet, InceptionResnetv2, MobileNetv2, and DenseNet201. Due to
their cheap storage needs, fast training times, and excellent accuracy, such techniques
should be combined with the utilization of ShuffleNet, SqueezeNet, and MobileNets. In
addition, there have been inadequate pest detection, classification, and counting
investigations considering abiotic factors. Some insect pests may exhibit unique
behaviors depending on abiotic elements and influencing artificial intelligence
performance. Consequently, additional research into these abiotic aspects is needed.
Many studies identified the pests but did not further estimate the insect population
dynamics.
Go to:

11. Conclusions

So far, AI models and IoT sensor-based techniques can identify, classify, and count
cotton insect pests or beneficial insects with reasonable accuracy. However, the main
challenges were the insect location (on plant leaves, stems, holes in soils), obtaining
enough data, identifying pests/beneficial insects from images with concentrated pests
and similar-looking species, or obtaining insects’ acoustic signatures containing noise.
On the other hand, IoT was limited by data storage, field remoteness, and a lack of
defined optimum distance between the sensors when targeting pests or beneficial
insects to estimate populations. Enhanced CNN algorithms, sticky boards, low-
bandwidth devices, and more research on insect-specific attractants to bring them to
the sensor region may overcome the mentioned challenges and improve AI and IoT in
cotton production. In addition, our analysis found that few studies on AI and IoT
explicitly address cotton pest predators/parasitoids and immature insects. Overall, this
review has identified the current pest classes already detected by AI, pest recognition
features, common variables, methods, intelligent systems, success, and challenges when
identifying, classifying, and counting pests or beneficial insects in cotton. By combining
these aspects, this review has contributed to the art of AI and IoT by identifying crucial
elements in the design and development of future devices to easily monitor and detect
pests/beneficial insects in cotton production.
Go to:

Acknowledgments

The authors would like to thank Yusto Y. Mgisha and Festo R. Silungwe for their
contributions and thoughts. We also appreciate Werenfrid Michael and Peter C.
Ngimbwa for their design, labeling, and suggestions. The authors would also like to
thank the University of Georgia in the United States and the University of Sokoine
Agriculture in Tanzania for their assistance throughout the review sessions.
Go to:

Nomenclature
CNN Convolutional Neural Networks
SVM Support Vector Machine
KNN K-Nearest Neighbors
ANN Artificial Neural networks
RNN Recurrent Neural Networks
DBN Deep Belief Network
DBM Deep Boltzmann Machine
LBP-SVM Local Binary Patterns with Support Vector Machine
Faster R-CNN Faster Recurrent Convolution Neural Network
ResNet Deep Residual Network
DCNN Deep Convolutional Neural Network
HD-CNN Hierarchical Deep Convolutional Neural Network
SegNet Semantic Segmentation Network
SSD MobileNet Single Shot Multi-Box Detector MobileNet
BP Neural Network Back Propagation Neural Networks
Bi-Directional
Bi-directional Recurrent Neural Network
RNN
LSTM Long Short-Term Memory
GNSS Geographical Navigation Satellite Systems
Go to:

Funding Statement

This review study was funded by USDA Non-Assistance Cooperative Agreement #58-
6080-9-006, Managing Whiteflies and Whitefly transmitted Viruses in Vegetable Crops
in the Southeastern US
 https://www.sciencedirect.com/science/article/pii/S258972172030012X

Artificial Intelligence in Agriculture

Volume 4, 2020, Pages 58-73

Implementation of artificial intelligence in agriculture for

optimisation of irrigation and application of pesticides and
herbicides
Tanha Talaviya a, Dhara Shah a, Nivedita Patel b, Hiteshri Yagnik c, Manan Shah d

a- Department of Computer Engineering, Indus University Ahmedabad, Gujarat, India

b-
Department of Computer Engineering, Nirma University Ahmedabad, Gujarat, India

Highlights
 • Authors have studied different automated systems employed in
agricultural sector.

 • AI is important for maintaining advancement in the agro-based sector.

 • Irrigation and weeding was mitigated by automated robotic systems.

 • The problems related to crop monitoring and spraying was solved.

Abstract
Agriculture plays a significant role in the economic sector. The
automation in agriculture is the main concern and the emerging subject
across the world. The population is increasing tremendously and with
this increase the demand of food and employment is also increasing. The
traditional methods which were used by the farmers, were not sufficient
enough to fulfill these requirements. Thus, new automated methods were
introduced. These new methods satisfied the food requirements and also
provided employment opportunities to billions of people. Artificial
Intelligence in agriculture has brought an agriculture revolution. This
technology has protected the crop yield from various factors like the
climate changes, population growth, employment issues and the food
security problems. This main concern of this paper is to audit the various
applications of Artificial intelligence in agriculture such as for irrigation,
weeding, spraying with the help of sensors and other means embedded in
robots and drones. These technologies saves the excess use of water,
pesticides, herbicides, maintains the fertility of the soil, also helps in the
efficient use of man power and elevate the productivity and improve the
quality. This paper surveys the work of many researchers to get a brief
overview about the current implementation of automation in agriculture,
the weeding systems through the robots and drones. The various soil
water sensing methods are discussed along with two automated weeding
techniques. The implementation of drones is discussed, the various
methods used by drones for spraying and crop-monitoring is also
discussed in this paper.
Table 2. Summary of applications of AI in weeding operations.
No Application Crop Algorithms for Weed Weed Removal Accuracy Reference
. Detection Methods
Pepper
Precision Weed plants, Machine Vision, Partel et al.
1. A smart sprayer –
Management artificial Artificial Intelligence (2019)
plants
High power
Autonomous Machine vision lasers for intra- Bakker et al.
2. Sugar beet –
Weeding Robot. algorithm row weeding (2006)
proposed.
Data augmentation for
Weeds
image preprocessing;
Detection in Ngo et al.
4. – Convolutional neural 70.5%
Agricultural Herbicide Spray (2019)
networks for weed
Fields
detection
Rotatory hoe/ Åstrand and
Robot for weed 92%
5. Sugar Beets Machine Vision Mechanical Baerveldt
control (detection)
removal (2002)
Motion of robot
Nakamura et
6. Weeding Robot Rice – prevents weed –
al. (2016)
growth
Weed Maruyama
7. Prevention Rice – Motion of robot – and Naruse
Robot (2014)
Color Based and Texture
Based algorithms; Robotic arms for
Sujaritha et
8. Weed Detection Sugarcane Greenness Identification; mechanical 92.9%
al. (2017)
Fuzzy Real Time removal
Classifier
Weed Control Electrical 84% Blasco et al.
9. Lettuce Machine Vision
System Discharge (detection) (2002)
Machine Vision
Robotic Weed algorithm based on Chemical Lamm et al.
10. Cotton 88.8% sprayed
Control Mathematical spraying (2002)
morphology
4. Drones in agriculture
Unmanned aeronautical vehicles (UAVs) or unmanned ethereal
frameworks (UAS), otherwise called automatons, in a mechanical setting
are unmanned aircrafts that can be remotely controlled (Mogli and
Deepak, 2018). They work in confluence with the GPS and others sensors
mounted on them. Drones are being implemented in agriculture for crop
health monitoring, irrigation equipment monitoring, weed identification,
herd and wildlife monitoring, and disaster management (Veroustraete,
2015; Ahirwar et al., 2019; Natu and Kulkarni, 2016). Remote
Sensing with the use of UAVs for image capturing, processing, and
analysis is making a huge impact on agriculture. (Abdullahi et al., 2015).
The rural business appears to have grasped ramble innovation with great
enthusiasm, utilizing these propelled instruments to change current
agricultural methods (Pederi and Cheporniuk, 2015). The complete
addressable estimation of automation fueled arrangements in every
single relevant industry is critical – more than USD 127 billion, as
indicated by an ongoing PwC analysis. They can be contrasted with a
normal simple to use camera for unmistakable pictures, yet while a
standard camera can give some data about plant development, inclusion
and different things, a multispectral sensor extends the utility of the
procedure and enables farmers to see things that can't be found in the
noticeable range, for example, moisture content in the soil, plant health
monitoring. These could help defeat the different restrictions that
obstruct agrarian production. The development of the UAS is
incorporated with Wireless Sensor Networks (WSN). The data recovered
by the WSN enables the UAS to advance their utilization for instance to
restrict its spraying of synthetic compounds to carefully assigned
regions. Since there are abrupt and continuous changes in ecological
conditions the control circle must almost certainly respond as fast as
could reasonably be expected. The reconciliation with WSN can help
toward that path (Costa et al., 2012). In precision agriculture, UAVs are
mainly applicable for agriculture operations such as soil and field
analysis (Primicerio et al., 2012), crop monitoring (Bendig et al., 2012),
crop height estimations (Anthony et al., 2014), pesticide Spraying (Faiçal
et al., 2017; Faiçal et al., 2014a, Faiçal et al., 2014b, Faiçal et al.,
2014c; Huang et al., 2009). (See Table 4.) However, their hardware
implementations (Maurya, 2015) are purely adherent on critical aspects
like weight, range of flight, payload, configuration and their costs. A
research involving technologies, methods, systems and limitations of
UAVs are examined (Huang et al., 2013). About more than 250 models
are analyzed as well as summarized in order to choose an appropriate
UAV in agriculture (S.R. Kurkute et al., 2018) (See Table 3.). The
agricultural drone market is expected to grow over 38% in coming years.
It is believed that the need for efficient agriculture is only going to
become more important due to increasing population levels and changing
climate patterns (Puri et al., 2017).
Table 4. Summary of various applications of drones in agriculture.
No Application Technologies/algorithms used Results Reference
.

Wireless Sensor Networks, Garre and

Pesticide
1. Gyroscope and Accelerometer N/A Harish
Spraying
sensors (2018)

Crop
UAVs could be used in order to detect
Monitoring, DJI Phantom 3 Advanced UAV Psirofonia et
2. abnormalities and identify potential
Mapping, and and other softwares al. (2017)
problems.
Spraying

Linear regressions between NDVI and

plant nitrogen, aerial biomass, etc. were
Vega et al.
3. Crop Monitoring Multispectral sensor significant. This has the potential to
(2015)
provide insight to good management
practices and techniques.

Pesticide Worked satisfactorily when tested on Yallappa et

4. Spray motor
Spraying groundnuts and paddy crops al. (2017)

………

Provides an approach for the

segregation of sparse and
dense areas within a
Precision Murugan et al.
– sugarcane field.
Agriculture (2017)
It makes use of satellite data.
Monitoring
Accuracy was 87% for
testing.
Spraying It has the ability
Pharne
Fertilizers Accelerometer and to reduce time
10. et al.
and Gyroscope Sensors, Arduino and human
(2018)
Pesticides effort.

4.1. Crop spraying

The UAVS, otherwise called drones, are chiefly established on the

innovations of sensors and microcontrollers which are grown especially
with an expectation to make up for the nonattendance of the pilot and
accordingly empower the trip of unmanned vehicles and their
independent conduct (Spoorthi et al., 2017). These drones have been
utilized as substance sprayers by farmers since numerous years now and
they are considered as effective and of great importance in the situations
of cloudy climate and has also solved the problem of inaccessibility to a
field of tall crops, for example, maize (Sugiura et al., 2005; Simelli and
Tsagaris, 2015). Additionally, they are likewise accepted to have a solid
favorable position contrasted with satellite airborne sensors of high
picture resolution (Jannoura et al., 2015; Simelli and Tsagaris,
2015). Giles et al., 1987 retrofitted an air-carrier plantation sprayer with
a microcomputer based sprayer control framework. A foliage volume
estimation framework, in view of ultrasonic range transducers was
interfaced to a PC which controlled the 3-nozzle manifolds on each side
of the sprayer by the utilization of control calculations dependent on the
amount of spray deposited.
4.2.1. Yield mapping and monitoring
One of the key segments of the unprecedenting progressions in exactness
cultivating frameworks, yield mapping, enables the farmer to see spatial
variety over the field perceiving zone for future activities and outcome of
the past sessions, management. It alludes for the most part to the way
toward gathering geo-referenced information on harvest yield and
qualities, for example, showing in-field fluctuation, and the soil moisture
content of the yield giving a benchmarking apparatus, when the yield is
being harvested. In combination with soil examining data, yield maps
empowers the arrangement of variable compost maps which considers
soil supplement levels just as the supplement which was expelled in the
collected harvest. Last result of yield mapping is typically a tonal or
shaded guide showing scopes of yield inside a field. Fundamental
segments of grain yield mapping framework incorporate grain fow sensor
(determines grain volume gathered), grain moisture content sensor
(remunerates for grain moisture variability), GPS antenna (receives
satellite sign), Yield screen show with a GPS receiver (geo-reference and
records information), header position sensor(distinguishes estimations
logged during turns), travel speed sensor (determines the separation the
join goes in a specific logging interim) (Fig. 3).

Its speculated that AI-based advisories would be useful to increase

production by 30%. The biggest challenge to farming is the crop damage
due to any kind of disasters including the pest attack. Most of the time
due to lack of the proper information farmers lose their crops. In this cyber age,
the technology would be useful for the farmers to protect their cultivation from
any kind of attacks. AI-enabled image recognition will be useful in this direction.
Many companies have implemented drones to monitor the production and to
identify any kind of pest attacks. A robotic lens zooms in on the yellow flower of
a tomato seedling. Images of the plant flow into an artificial intelligence
algorithm that predicts precisely how long it will take for the blossom to
become a ripe tomato ready for picking, packing, and the produce section of a
grocery store. The technology is being developed and researched at
NatureFresh Farms, a 20-year-old company growing vegetables on 185 acres
between Ontario and Ohio. Knowing exactly how many tomatoes will be
available to sell in the future makes the job of the sales team easier and directly
benefits the bottom line, said Keith Bradley, IT Manager for NatureFresh
Farms. It's only one example of AI transforming agriculture, an emerging trend
that will help spur an agricultural revolution. Artificial intelligence can help
humanity confront one of its biggest challenges: feeding an additional 2 billion
people by 2050, even as climate change disrupts growing seasons, turns arable
land into deserts, and floods once-fertile deltas with seawater.

The United Nations estimates we will need to increase food production 50% by
the middle of the century. Agricultural production tripled between 1960 and
2015 as the world's population grew from 3 billion people to 7 billion. While
technology played a role in the form of pesticides, fertilizers, and machines,
 much of the gains can be attributed to simply plowing more land—cutting
forests and diverting fresh water to fields, orchards, and rice paddies. We will
have to be more resourceful this time around. AI is likely to transform
agriculture and the market in the next few years. The technology has been
useful for the farmers to understand various types of hybrid cultivations which
would yield them more income within the limited time frame. The proper
implementation of AI in agriculture will help the cultivation process and to
create an ambiance for the market. There is a huge wastage of the food across
the world and using the right algorithms, this problem can also be addressed.

https://arccjournals.com/journal/agricultural-reviews/R-2296

Role of Artificial Intelligence in Indian Agriculture: A

Review
Vinod Chandra S.S.1,* Emailvinod@keralauniversity.ac.in, Agricultural
Reviews, Volume 44 Issue 4 (December 2023) : 558-562

Affiliations
Department of Computer Science, University of Kerala, Thiruvananthapuram-695
1

034, Kerala, India.

Cite article:- S.S. Chandra Vinod (2023). Role of Artificial Intelligence in Indian
Agriculture: A Review . Agricultural Reviews. 44(4): 558-562. doi: 10.18805/ag.R-
2296.

ABSTRACT
The next generation is very keen on agriculture and is looking forward to high
yield productivity by using the latest technologies. Artificial intelligence tops the
list as it creates an eco-system that is favourable to all stakeholders involved.
Technology should produce a framework that helps the farmers in every aspect
of cultivation, starting from land preparation to selecting seeds to fertilize. The
use of technology can help them to identify the apt plantations for each season.
Market analysis and user demands can also be foreseen by the farmers, which
helps them get a better yield. Artificial intelligence-based farming creates high
productivity and better yield, which increases the profit of individual farmers. An
end-end market where the customer gets quality products directly from the
producers increases the quality and reduces the price hike. Technology-driven
farming will pull the next generation to take up agricultural jobs shortly. This
method will help farmers to understand the demand of the market through
prediction forecasting. Artificial intelligence and its allied fields are tuned to ‘agro
computer’ in tomorrow’s world.

KEYWORDS
 Agriculture
 Agro-Computer
 Agro-Market
 Artificial Intelligence
 Sos Framework

INTRODUCTION
Agricultcre is the backbone of our country's economy and nearly 70-75% depend
directly or indirectly on it. Even in a country like ours, it is hard to believe that
methods and markets of agriculture are way low the basic standards. There are
no proper guidelines available for farming. No proper market study and not even
scientific farming methods are reaching the farmers. Markets and technologies
are never with the producers and they are not even aware of new farming
methods. Israel is a very good example where advanced farming methods are
employed for high yielding. Here in India, not even our agricultural research
institutes are developing technologies that help agriculture.

Let's see how artificial intelligence and expert systems can contribute to the
country's agricultural development. When talking about intelligence, human
intelligence is one of the most complex systems. The two aspects of intelligence
are learning and dealing with things and the ability to rationalize thought.
Artificial intelligence is a computer science branch with many applications in
different fields (Vinod and Anand, 2020). It learns how to program computers to
behave intelligently. It has to do with a job similar to using computers to
understand human intelligence.

What are the use and use of artificial intelligence for the common person? Of
course, in today's world, the use of artificial intelligence is useful. Although
difficult for the common person to understand, artificial intelligence profoundly
affects everyone in society and plays a vital role in progressive modern life. High
tech agriculture tools, smart transporting and consistent field analytics to
farmers and insurers will help agriculture bloom. In digital framing, high yield
productivity is expected. It applies precision location methods and decision
quality agronomic information to brighten, predict and affect the range of
cultivation issues across the farm.

Deploying science and technology in agriculture is essential to eradicate severe

poverty and famine. As part of a study, Microsoft Corporation has been operating
in Andhra Pradesh, India, since 2018, providing services and solutions to 175
farmers (Gurumurthy and Bharthur, 2019). In collaboration with International
Crops Research Institute for the Semi-Arid Tropics (ICRISAT), Microsoft developed
an artificial intelligence sowing App that uses machine learning and business
intelligence from the Microsoft Cortana Intelligence Suite. It uses powerful cloud-
based predictive analytics to empower farmers with crucial information and
insights to reduce crop failure, increase yield, reduce stress and make better
income (Diksha, 2020). It involves land preparation, sowing of seeds, climate
change, fertilizer addition, software application, organic fertilizers and other
nutrients for the crop. On average, the yield per hectare has increased by 30%
compared to the previous harvest.
Predictive analytics will help the framers for timely crop sowing, make
predictions on the right time to plant, apply fertilizers, harvest, etc., based on
climate data, historical conditions, market conditions, personal information, etc.
With full knowledge and consent, digital technologies can have conquered the
agricultural sector. But right now, we are in a phase where artificial intelligence
is entering the Indian agricultural sector without much groundwork. This will
badly affect our farmers and markets. Mainly our agricultural community,
primarily farmers and those who depend on them. The second impact will be on
the agricultural market and those who depend on it. Access to the market
through unreliable market connectivity, fluctuations in commodity prices, low
income and high debt make agriculture a high-risk factor. In the nineteenth
century, during the Industrial Revolution, machines were deployed to reduce
human resources. Now in this twenty-first century, demands are for modern
methods and technological updations used in agriculture.

The primary experimental study was conducted during 2018-2020 along with
Kerala Agricultural University and the Department of Computer Science,
University of Kerala. The review materials were selected by the help of
agricultural scientists of the Kerala Agricultural University.

Technologies in agriculture

Artificial intelligence and IoT based remote access micro-grid power farming has
slowly started in India. Indian government organizations like C-DAC is providing
facilities to farmers aimed at promoting agriculture. New technologies need to be
implemented as the demand for agricultural products has increased over time.
Crop productivity has increased over time as a result of advancements in
information technology, which resulted in producing field yield seeds. The use of
computers played a significant role in heavy crop production in the twentieth
century. There is no doubt that artificial intelligence will play a role in the next
decades. It is a fact that artificial intelligence offers ways to facilitate an increase
in farmers' incomes, a rise in agricultural productivity and even waste reduction.
Artificial intelligence plays a significant role in all these areas, thus preventing
"agriculture as a service" from being a fragile segment. The uses of cognition are
spread over eight main areas. This will benefit the Indian agricultural sector.
They need to come up with technologies, especially artificial intelligence.
1. Internet of Things (IoT) led growth.
2. Image-based insight work.
3. Prepare suitable mixtures for agricultural products.
4. Strengthen crop health monitoring
5. Enable farmers to use irrigation and its maximum use.
6. Implement self-sustaining technology projects in agriculture.
7. Direct value-added products and manufacturing and enterprise marketing in
the right direction.
8. Understand the right direction of the market.

The demand for data-driven framing and predictive analytics in agriculture is a

guide after the post-COVID 19 worlds. The framers and agro-business owners
are actively taking up precision farming methods supported by artificial
intelligence and machine learning frameworks. These frameworks support
farmers in crop and livestock management by removing the element of
guesswork, forecasting yield, managing a supply chain network and assessing
risks (Saxena et al., 2020). Blockchain technology gives a transparent supply
chain in various agriculture fields, starting from seed identification to harvest. It
will provide popularity among the framers and systems regarding technical
aspects, education, policies and regulatory frameworks.

Challenges in adopting technologies in agriculture

The Niti Aayog, Government of India, recently published a discussion

paper (Tanay, 2019). It envisions solutions using artificial intelligence for key
areas, including agriculture, industrialization, etc. In agriculture, technologically
advanced machinery and built-in intelligence give farmers knowledge about soil
quality, when to sow, spray herbicides, where pests occur, etc. If we built an
intelligent system, then it can advise farmers on various best practices, India
thus can see a new agricultural revolution. However, there is a strong challenge
for that futuristic situation. The entire supply chain uses capacity expansion and
cost reduction factors, which can backstab the Indian farming population. Even
though built-in intelligence-based technologies have unique advantages, there
are several challenges in adopting them in the agricultural sector (Lal Mohan et
al., 2019).

1. Reliability of system and technology.

2. Security and acceptance of information.
3. Data privacy and storage and its use.
4. Social acceptance and recognition.
5. Live release, accessibility and use of reliable information.
6. Cost-effectiveness.
7. Ease of use and training.
8. Unethical stakeholders.

We must realize the realities of Indian agriculture and its market. Farmers are
dependent on various external factors for agriculture and harvest and many a
time, they are not guaranteed, which makes farming a high-risk activity. Even
the market plays a vital role; since the products are mostly perishable, they have
to accept whatever the market offers. Lack of quality storage facilities also is a
reason why the farmers are getting exploited by the market. Certain agro-
products like milk, egg, meat and vegetables will be damaged after a limited
time. If artificial intelligence-based techniques were deployed, it would have
given optimum solutions in cold storage, transportation and demand for these
products. For example, every place has its festivals. The market need of that
time is different and if it can be predicted, then the demand and supply could be
foreseen and a new market opportunity could be fetched.
Application of technology farmers in India have a very small landholding and
they are unable to sustain the cost of buying the seeds and other essentials.
Here, artificial intelligence for predicting the weather at a local level guides
farmers to use sustainable techniques (Tanha et al, 2020). This helps framers
manage pests through ecology, robots for harvesting in a multi-crop farm,
demand prediction available stocks, export, local needs, etc.

The importance of agriculture demands in the world increases because of impact

issues after COVID 19. Technological advancements in agriculture farming and
its marketing lead to International economic growth and the prevailing lockdown
situation worldwide. It further resulted in food supply chain disruptions using
predictive techniques.

Artificial intelligence models for agriculture

Globally, built-in technologies are emerging to improve the efficiency of

agriculture, seed quality, soil monitoring, climate forecasting, forecast
agricultural analysis, markets and distribution networks. We can use cloud
computing infrastructure, data eco-systems, the Internet of Things (IoT) and
machine learning to make more productive farming. We can focus on the
development of the digital agricultural sector. For this, focus on efficient farming,
efficient irrigation, smart chemical spraying, effective harvesting and pest
control. Artificial intelligence-based automated irrigation dramatically increases
the efficiency of the current agriculture supply chain. The percentage of
freshwater resource requirements increased with the increase in food demand.
Hence, artificial intelligence plays an importance in smart irrigation, water level,
soil temperature, nutrient content and weather forecasting (Tanha et al.,
2020). Models can be created to achieve such demands. Machine learning and
soft computing techniques, a subfield of artificial intelligence, can be adapted for
model building. Use of plant growth images/videos, we can make intelligence
models using the above techniques. Built-in intelligence is widely used worldwide
for drone cameras, satellite imagery, data processing, systems for monitoring
and managing various agricultural activities, disease and pest detection, weather
forecasts and fertilizer application times. A sub-branch of artificial intelligence
can predict the proper use of chemical sprays, harvest time and product life
through machine learning. For example, the harvest time and weight of bananas
grown on the same soil on the same day are different. This can happen even if
the fertilizer is applied evenly. Such problems can be technically examined and
resolved. The complexity increases when multiple such farming problems are
combined. These problems can be solved by building and using specialized built-
in intelligence models with the necessary learning. In-depth study methods using
agricultural data, crop diagrams, growth rates and disease data provide useful
forecasting assumptions on market and harvest (Hatim et al., 2019).

Today banking and financial institutions give artificial intelligence-based lending

solutions to the farmers. The payback capacity can identify before sectioning the
amount using the data analytics by machine learning. Similarly, insurance
companies are using machine learning frameworks to augment the processes
related to the current crop insurance schemes, preventing the claimed delay and
reducing the timeline in claim settlements.

Robots and intelligent machines are useful for farming and harvesting, leading to
huge savings by reducing need for agricultural laborers. Identify the right seeds'
availability for the next seed germination and flowering machine learning
framework for Indian farmers. For example, many seed banks in India need to
coordinate with a machine learning framework for distribution to the framers.
The seed suggestions are based on that farming place's soil structure,
availability, market and weather conditions.

Artificial intelligence-based framework

The Government's Economic Survey 2020 identified the data-based facility as a

critical characteristic (Arka, 2018). It underscored the desire to build things
based on people's data. Agriculture and the agricultural market focused on it.
This is where the open potential for the study of our artificial intelligence in
agriculture lies. For making a framework, acceptable working practices are
required. Through our research, agriculture and farming methods will benefit the
country as a development model under artificial intelligence supervision. This is
being implemented in other developed countries.

A model framework (systems of sedentary behaviours or SoS) is shown in Fig 1.

The Indian monsoon rainfall is predicted using artificial intelligence techniques
such as artificial neural networks. Indian meteorological department has
developed rainfall and climate change forecasting methods. Including these
technologies into the SoS framework, the framers get a maximum yield and
agricultural profit. This can be done in many ways. In this way, a sense of
direction can be given to the next generation. Natural language processing
(NLP), robotics, machine learning automatic reasoning, knowledge
representation, expert systems, computer vision, speech recognition, automated
data analytics and cloud computing, statistical computing and deep learning are
some of the key subfields of artificial intelligence that can be used to solve
complex problems in agriculture (Konstantinos et al., 2018). Many of these are
used by our farmers through technological software. For example, a stand-alone
Q&A system in Indian languages for farmers. The deep learning-based
agriculture mapping from satellite images gives an intelligent forecasting system
to the farmers and agro-business. It is built on top of an image data stream to
help recognize soil science, crop yield, pest management, harvesting, etc.
Fig 1: Artificial intelligence-based SoS.

Precision agriculture focused on improving the overall harvest quality and

accuracy through artificial intelligence methods. The detection of target weeds
can be by artificial intelligence sensors and then routed to herbicide region
selection. This is by automation mechanism to the agriculture farming. The
storage of seeds for the next season is one of the challenging tasks. Using data
analytics by machine learning techniques provide a clear insight to the
distributors. Similarly, the storage and distribution of grains to the market and
users’ need an intelligent forecasting technique. Forecast the price of agro-
products in the present and futures markets are today’s requirements. Also,
strengthening the health of crops is challenging. An artificial intelligence-based
framework will help those farmers by suggestions. Crop diseases are a severe
problem that leads to low yield, productivity and unrespecting causes to the local
farmers (Selvaraj et al., 2019). Better identification of crop disease in advance
using machine learning techniques offers remedial solutions to high productivity
farmers.

Challenge of Artificial Intelligence Research

One of the major challenges is how artificial intelligence research is carried out in
our country. Current research and development initiatives are concentrated in
top Indian institutions such as the Indian Institute of Technology (I.I.T), National
Institute of Technology (N.I.T), etc. Here a community of 50 to 75 people works in
this domain, and with this small workforce, they cannot pull off sufficient
advancements. Thus, India lags far behind in bringing high-quality outputs in
smart and intelligent computation. Space, lack of resource and administrative
barriers, approach to research, poor computing facilities, lack of interpreted
quality data have been cited as major pain affecting India's artificial intelligence
research. The increase in Digital India's budget to US $ 477 million in 2018 to
expand artificial intelligence research was a welcoming first step. But even with
these funds, addressing institutional weaknesses continue to be a major
challenge. It is also not clear how public agricultural universities and research
institutes will benefit from such allocation. Artificial intelligence applications in
the Indian agricultural sector need to grow, and the technologies need to be
brought down to the root level. All stakeholders should be benefitted.
Technologies should never remain in laboratories, and it should be handed down
and tested in reality. Recommendation systems will benefit consumers and
farmers in identifying the entire supply chain, which will quickly benefit both
parties. If we adopt artificial intelligence-driven methodologies at different
farming stages, we can witness an era of flourishing agricultural resources. This
can only be achieved by advancing multiple research methods simultaneously.

CONCLUSION
Today's world is moving to next-generation agriculture by using artificial
intelligence framework, which can be termed as smart agriculture system. For
better yield in agriculture, it is recommended,
1. Increase support for research and development activities in the area of
artificial intelligence in agriculture.
2. Implementing expert guidance strategies, large scale research and
application units.
3. Effective evaluation of equipment using artificial intelligence and related
methodologies.
4. Improve interaction with all stakeholders at multiple levels of the supply
chain.
5. Provision for agro educational research efforts.
6. A holistic system approach enables scalability and rapid response to changes
in water, climate and employment availability, i.e., adopting an SoS framework.
7. Development of secure, adaptive and cohesive systems.
8. Understand the innovative and intelligent processes that take place at
various agricultural universities and their impact on agriculture.
9. Understand the term "agricultural computer" on an experimental basis. That
is, see the interaction of artificial intelligence and sub-sectors in the agricultural
domain.
10. Adopt an integrated, multi-dimensional research approach in agriculture.

CONFLICT OF INTEREST
• All authors have participated in (a) conception and design, or analysis and
interpretation of the data; (b) drafting the article or revising it critically for
important intellectual content and (c) approval of the final version.
• This manuscript has not been submitted to, nor is under review at, another
journal or other publishing venue.
 • The authors have no affiliation with any organization with a direct or indirect
financial interest in the subject matter discussed in the manuscript.

REFERENCES
1. Arka Bagchi (2018). Artificial Intelligence in Agriculture, Mindtree White Paper,
Available from https://www.mindtree.com/sites/ default/files/2018-04/Artificial
%20Intelligence%20in%20 Agriculture.pdf.
2. Diksha, M. (2020). Artificial Intelligence: A new way to Improve Indian
Agriculture. International Journal of Current Microbiology and Applied Sciences.
9(3): 1095-1102. DOI: 10.20546/ijcmas.2020.903.128.

3. Gurumurthy, A. and Bharthur, D. (2019). Taking Stock of Artificial Intelligence in

Indian Agriculture. Friedrich-Ebert-Stiftung, DOI:10.13140/RG.2.2.13914.36801.
4. Hatim, G., Lara, P., Jennifer, R. and Son, C. (2019). Climate adaptive smart
systems for future agricultural and rangeland production. White Paper,
https://aces-cesfas.nmsu.edu/ documents/artificial-intelligence-in-agriculture---
white- paper-with-summary---v6.pdf.
5. Konstantinos Liakos, Patrizia Busato, Dimitrios Moshou, Simon Pearson and
Dionysis Bochtis (2018). Machine learning in agriculture: A review. Sensors in
Agriculture. 18: 2674. DOI: doi.org/10.3390/s18082674.
6. Lal, M.B, Ramasubramanian, V., Arora, A., Marwaha, S. and Parsad, R. (2019). Era
of Artificial Intelligence: Prospects for Indian Agriculture. Krishi. 69(3): 10-13.
DOI: krishi.icar.gov.in/jspui/handle/123456789/19207.
7. Selvaraj, M., Alejandro, V., Henry, R., Nancy, S., Sivalingam, E., Walter, O. and
Guy, B. (2019). AI-powered banana diseases and pest detection. Plant Methods.
15: 92. DOI: doi.org/10.1186/s13007-019-0475-z.
8. Saxena, A., Suna, T. and Saha, D. (2020). Application of Artificial Intelligence in
Indian Agriculture, RCA Alumni Association, India.
9. Tanay, M. (2019). Role of Digital and AI Technologies in Indian Agriculture:
Potential and Way Forward, NITI Aayog, September (2019).
10.Tanha, T., Dhara, S., Patel, N., Yagnik, H. and Shahd, M. (2020). Implementation
of artificial intelligence in agriculture for optimization of irrigation and application
of pesticides and herbicides. Artificial Intelligence in Agriculture. 4: 58-73. DOI:
10.1016/j.aiia.2020.04.002.
11.Vinod Chandra, S.S., Anand, H.S. (2020). Artificial Intelligence: Principles and
Applications, Printice Hall India, pp.428.
 https://epubs.icar.org.in/index.php/IJPP/article/view/135084

Mega-field demonstration of management of pink

bollworm, Pectinophora gossypiella in rain-fed cotton
using mating disruption technology

C D Mayee 1Agrovision Foundation, Nagpur, 2 South Asia Biotechnology

Center, Jodhpur-New Delhi

 B Chaudhary South Asia Biotechnology Center, Jodhpur-New Delhi

 R Panchbhai Agrovision Foundation, Nagpur
 A R Annepu PI Industries, Gurugram, Haryana
 R D Kapur PI Industries, Gurugram, Haryana

Keywords:
Bt cotton, mega demonstration, Pectinophora gossypiella, pink bollworm, PBKNOT
dispenser, PBW-Management

Abstract
With large scale outbreak of pink bollworm, Pectinophora gossypiella on cotton since
last four years, the benefits of Bt cotton in terms of saving the chemical usage are
waning. There are reports of resistance break down of cry toxins as also of some
insecticides on multiple Bt cotton hybrids which occupy nearly 95% of the total
cotton area. Efforts to educate the farmers through mass communication techniques
on do’s and don’ts have only partially been beneficial as the use of insecticides
become the part of package of management. Therefore, non-chemical approach of
mating disruption technology was demonstrated in the present study as an area-
wide approach in cotton fields of farmers. A mega demonstration, called ‘Project
Bandhan’, on 300 acres in five cluster of 60 acres each in three villages of Nagpur
district of Maharashtra was organized during the rainy season of 2021-22. The mega-
demo was carried out using the latest pheromone based PBKNOT dispensers
supplied by PI Industries Ltd., Gurugram, Haryana. The PBKNOT dispenser was tied
on to main stem at 45-50 days after planting. The results revealed that PBKNOT tied
fields reduced the flower damage by 49.2%, boll damage by 58.3% and locule
damage by 51.8% averaged over 300 acres of treated clusters. At the end of 10th
week, the active ingredient of PBKNOT was still active as more than 93% mating
disruption was visible indicating the slow release of pheromone even after three
months. The average yield increase based on 300 acres was to the tune of 550
kg/ha. Being a large-scale demonstration, it appears that the technology is good
option for the rain-fed cotton farmers to reduce the use of chemicals for
management of pink bollworm in group farming system.
 https://www.researchgate.net/publication/
344273586_Assessment_of_bollworm_damage_and_yield_loss_in_seed_blends_of
_Bollgard-II_with_corresponding_Non-Bt_hybrid_as_'built_in_refuge'_in_cotton

Assessment of bollworm damage and yield loss in

seed blends of Bollgard-II with corresponding
Non-Bt hybrid as ‘built in refuge’ in cotton
 April 2021 Phytoparasitica 49(06):1-11

 Rishi Kumar
 S. Kranthi

 gmv Prasada Rao

 Show all 8 authors

 Keshav Kranthi

Abstract and figures

Bt cotton plants, with bollworm resistant trait, have been widely adopted since
introduced in India in 2002. As a proactive measure to prevent evolution of
resistance to Bt protein(s), it was approved for commercial cultivation with a
stipulation of structured refuge of minimum 20% of the Bt area, constituted by plants
that do not produce Bt toxins and thus allow growth and development of population
of Bt susceptible pests. Though the refuge cotton seeds are provided in a separate
pouch within each Bt cotton seed packet, Indian farmers have been reluctant to sow
refuge. To find out a method for ensured refuge planting, we experimented with
‘built-in-refuge’ (BIR) in which non-Bt cotton seeds were blended with Bt cotton
seeds in various proportions ie 5,10, 50% along with 20% recommended structured
refuge and each treatment was evaluated in terms of fruiting body damage by major
cotton lepidopterans, relative to the regulator stipulated 20% non-Bt structured
refuge at five locations during cotton growing season of 2012–13. A major outcome
from this study was that treatments with 5 and 10% non-Bt blends with bollga0rd-II
were as effective as the 20% structured refuge, based on fruiting body damage due
to bollworms, a key factor in managing Bt resistance without any compromise to the
seed cotton yield. It can be concluded that the seed blend technology is a good
option to impose refugia compliance to delay resistance in Bt cotton to bollworms in
a country having lesser alternate hosts available for the target bollworms during the
season and voluntary compliance of refuge is absent
 https://www.researchgate.net/publication/
357665924_Indian_scenario_on_the_occurrence_of_a_dreaded_insect_pest_Pink
_bollworm_Pectinophora_gossypiella_on_Bt_cotton-A_review

Indian scenario on the occurrence of a dreaded

insect pest Pink bollworm, Pectinophora
gossypiella on Bt cotton-A review
 January 2022
 Journal of Environmental Biology 43(1):11-19

 gmv Prasada Rao

View more

Abstract and figures

The pink bollworm, Pectinophora gossypiella, has become a significant production

constraint on Bt cotton in India. This problem is unique to India because the pest has
developed multi-fold resistance to Cry toxins in many Indian populations but not in
other countries. Most Indian populations have developed multifold resistance to Cry
1 Ac and Cry 1Ac + Cry 2 Ab toxins. Year-round cultivation of long-duration Bt cotton
hybrids on a large scale has a pronounced impact on the incidence. Also discussed
other factors responsible for the occurrence of pink bollworm on Bt cotton in India.
Insecticide Resistance Management (IRM) strategies implemented by different
cotton-growing countries globally; the USA, India, and China had a significant impact
on the interaction of pink bollworm on Bt cotton. Huge selection pressure resulted in
resistance to Cry toxins. Time-tested IPM, if implemented on a community basis
focusing on pheromones technology and closed season, will help sustain the cotton
cultivation in India in the future. Thus, this review aims to congregate exhaustive
information on the history, biology, resistance to Bt cotton, and Integrated
Management (IPM) options for the Indian scenario, which would help researchers in
their future endeavors.
 DOI: 10.1007/978-3-030-64504-5_16

 In book: Cotton Precision Breeding

https://www.researchgate.net/publication/
352235561_Critical_Evaluation_of_GM_Cotton

Critical Evaluation of GM Cotton

 DOI: 10.1007/978-3-030-64504-5_16

 In book: Cotton Precision Breeding

 June 2021

 Ganesan Balasubramani Kp Raghavendra Joy Das S kranthi

 Show all 8 authors
 K. R. Kranthi

Abstract

Over the past 24 years, biotech cotton has been influencing global cotton production
systems. Genetically modified (GM) cotton or biotech cotton has thus far been
approved in 19 countries. In 2018, biotech cotton was cultivated in 24.9 million
hectares, which comprised 76.0% of the global cotton area. More than 94.0% of the
global biotech cotton area is located only in five countries: India, the USA, Pakistan,
China, and Brazil. Thus far only two traits, namely, insect resistance and herbicide
tolerance (HT), are commercially available in GM cotton. Herbicide-tolerant cotton
has been developed to survive six different herbicides, namely, glyphosate,
glufosinate, dicamba, 2,4-D, isoxaflutole, and bromoxynil (HT not in use). Insect-
resistant cotton has been based mainly on seven Bt toxins, cry1Ac, cry1Ab, cry1C,
cry1F, cry2Ab, cry2Ae, and vip3a expressed by genes derived from the soil
bacterium Bacillus thuringiensis (Bt). Though cowpea trypsin inhibitor (CpTi) protein
was also deployed in insect-resistant biotech cotton in China, the area under
cultivation appears to be negligible. Biotech cotton is available in different forms with
either a single gene trait or with combinations of genes and traits. Initially Bt cotton
was commercialized with a single gene cry1Ac. The second-generation biotech
cotton contained cry1Ac + cry2Abcry2Ab, cry1Ac + cry1F, cry1Ab + cry2Ae, cry1 +
cry2 + vip3A, and Bt + HT (epsps). Bt cotton and HT cotton are grown in Argentina,
Australia, Brazil, Colombia, Mexico, Paraguay, South Africa, and the USA, while
developing countries such as India, Burkina Faso, China, Pakistan, Myanmar, and
Sudan approved only Bt cotton and have not approved HT cotton as yet. Two
bollworm species, namely, Helicoverpa zea (in the USA) and the pink bollworm
Pectinophora gossypiella (in India), have developed resistance to cry2Abcry2Ab and
cry1Actoxins in Bt cotton. Glyphosate resistance was recorded in 13 weed species
each in the USA and Australia and 8 each in Argentina and Brazil. Reports indicate
substantial economic benefits due to the adoption of GM cotton. Economic benefits
from Bt cotton can arise from higher yields due to effective protection from
lepidopteran larval damage and from savings due to reduced insecticides for
bollworm control. Research reports show that adoption of Bt cotton resulted in an
initial decrease of insecticide usage followed by an increase in some countries.
Insecticide usage has been increasing constantly over the past 10 years in India,
Pakistan, China, Brazil, and the USA for the control of thrips, whiteflies, mealybugs,
boll weevils, and pink bollworms. Insecticide use for boll weevil control is a major
concern in Brazil. Enhanced use of herbicides to control resistant weeds in the USA
and Brazil is an emerging concern. GM technologies so far have assisted growers in
managing bollworms and weeds. Target insects and weeds have been developing
resistance, thereby necessitating the need for new genes or new technologies. New
cotton varieties have been developed using CRISPR and RNAi and offer promise in
combating weeds, insect pests, diseases, and drought.

Balasubramani, G. et al. (2021). Critical Evaluation of GM Cotton. In: Rahman,

Mu., Zafar, Y., Zhang, T. (eds) Cotton Precision Breeding. Springer, Cham.
https://doi.org/10.1007/978-3-030-64504-5_16
 https://www.downtoearth.org.in/news/agriculture/cotton-curse-pheromone-
based-mating-disruption-technology-may-help-control-pink-bollworm-92547

Cotton Curse: Pheromone-based mating disruption technology may

help control pink bollworm

Experiments show 90% reduction in losses, improved yield
By Himanshu Nitnaware
Published: Monday 30 October 2023

This is the seventh story in a series about pink

bollworm attacks on Bt Cotton in the North Zone,
comprising Punjab, Haryana, and Rajasthan.
Farmers across cotton-growing states are struggling to manage the pink
bollworm (PBW) pest that is wreaking havoc, resulting in heavy economic
losses. Even the genetically modified pest-resistant cotton variety, Bt
Cotton (Bollgard II seed), which was created to resist PBW, is no longer
effective in keeping them away.

Ground visits by Down To Earth (DTE) revealed farmers in Haryana,

Rajasthan and Punjab were facing yield losses of 80-90 per cent.

PBW is a worm that damages parts of the developing cotton fruit by laying
eggs on it. The larvae burrow into the rounded sac of seeds called bolls to
feed on the seeds. The pest is concealed inside the green boll and is
protected; thus, spraying of pesticides and insecticides does not affect it
after it penetrates the fruit.

Read more: Untangling India’s Bt cotton fraud

YG Prasad, director, ICAR-Central Institute for Cotton Research, Nagpur

told DTE that scientists are working to identify new genes effective
against PBW.
“We have observed 8,000 germplasm accessions in field evaluations to
identify sources of natural resistance that will help fight PBW, but it is
likely that a transgene or gene editing will have to be tried,” Prasad said.

Even after the gene is identified and a seed variety is developed, it would
still take time to undergo various trials, create a public-private partnership
and seek approval from the Review Committee on Genetic Manipulation
and Genetic Engineering Appraisal Committee, he added.

“Effective monitoring and timely adoption of a crop-window-based

integrated pest management (IPM) strategy is the immediate solution,”
Prasad suggested as the immediate solution.

However, a group of scientists is experimenting with a technology that

has proven to be successful in the United States. The technology is called
PBKnot or PB Rope L, said Bhagirath Choudhary, founder-director for
South Asia Biotechnology Centre, Jodhpur, Rajasthan.

“The technology helped the US tackle the pest that was affecting farmers
for 100 years. The pest was eradicated in 2018 after its implementation in
2008,” he said.

Read more: Why are pests developing resistance to Bt cotton in

Telangana, Andhra Pradesh?

According to Choudhary, PBW annually affects lint quality and lowers yield
on about 12 million hectares of cotton crops worldwide. The Centre has
been experimenting with technology in India since 2022 and has received
enormous success, resulting in a 90 per cent reduction in losses, he said.

The technology received for the IPM strategies for controlling the
population of PBW received approval in India from Central Insecticide
Board and Registration Committee for the first time in 2019-20.

“The technology is being experimented with under Project Bandhan and

working closely with Krishi Vigyan Kendra, agriculture universities and
local institutions. It involves disrupting the mating cycle of the moths,” he
said.

The technology involves using a 20-centimetre polyethylene hollow rope

that is impregnated with pheromone gossyplure — a chemical emitted by
female moths — to attract male moth.

In natural temperatures, the rope expands and the narrow holes emit
pheromones, which released in the air, confuse the male moths and
prevent them from reaching the actual female moths, said Choudhary.
This, thus, disrupts the mating process and reproductive cycle. The result
is a drop in population and reduced damage to crops.

“They are installed after the crop is 45 days old and tied to plants around
the border and inside. About 160 such ropes are installed in an acre along
the border of a farm and inside with a gap of 25 square metres. The
technology is successful on large cotton fields with a minimum area of 40-
50 acres,” he said.

Read more: Study points to quality issues in Bt cotton seeds

A pheromone trap is also installed to monitor the activity of male moths,

along with checking flowering and green bolls. The method successfully
helped assist 90 per cent disruption in 1,100 acres of land across 18
locations in India in 2022, he said.

“It also helped increase yield up to 1.5 quintals per acre with improved lint
quality. In 2023, the experiment was being conducted in 710 acres of land
in 11 locations in Gujarat, Maharashtra, Telangana and Andhra Pradesh,”
the scientist said.

The government of India has to step in to adopt this biotechnology tool

and implement it in all cotton-producing zones in India to make it
successful on a wide scale, Choudhary added.
 https://www.downtoearth.org.in/news/agriculture/cotton-curse-tired-of-
losses-farmers-giving-up-cotton-on-a-large-scale-92498

Cotton Curse: Tired of losses, farmers giving up cotton on a large

scale
The production of cotton in Punjab has almost halved in the past decade
By Himanshu Nitnaware
Published: Friday 27 October 2023

This is the sixth story in a series about pink bollworm attacks on Bt Cotton
in the North Zone, comprising Punjab, Haryana, and Rajasthan.

Cotton farmers in the northern cotton zone have experienced the worst
pink bollworm attack on their crops in two decades this year; however,
pest attacks on cotton crops have occurred on a regular basis in recent
years. Farmers are being forced to switch to other crops due to continued
losses and uncertainty.

Ground visits by Down To Earth (DTE) to Punjab and Haryana have

revealed that many farmers are experimenting with horticulture and
paddy as an alternative to cotton farming. The landscape that once
displayed white cotton carpet stretched for kilometres is now being
replaced by paddy and orange plantations.

In Khariya village of Sirsa district in Haryana, Rajesh Nain has

experimented with growing paddy on 1.5 acres of land. “I planted cotton
on 20 acres of land and 90 per cent of it is lost to pink bollworm. The
paddy demands more water than cotton, which means exploiting more
groundwater, but if successful, I will switch to paddy altogether,” he said.

Read more: Untangling India’s Bt cotton fraud

Paddy cultivation requires less labour and has reduced risks of infestation,
said another farmer, Navin Surendrakumar, from the same village. “Even
if the quality is not optimal, it will sell for a good price,” he explained.

Cotton has a higher minimum support price (MSP), according to

Surendrakumar. “Paddy fetches around Rs 3,200, but it is higher if you
sow basmati varieties,” he said. The MSP for cotton is Rs 6,235 per
quintal.

But despite the lower price in the market, the investment cost of
pesticides and labour costs are far less for paddy, he pointed out.

Ranvir Kumar of Maujagarh, Punjab, converted his 20-acre cotton farm

into an orange orchard. The farmer suffered cotton crop losses in 2021 as
a result of pink bollworm and whitefly attacks.

Growing an orange orchard, according to the 32-year-old, is low-

maintenance. As a result, investing in the crop appears more reliable,
whereas growing cotton involves significant risks and is similar to
gambling.

Related Stories

 Is Bt cotton a success or failure?

 Farmers shun cotton
 Karnataka farmers upset over Bt cotton

Read more: Why are pests developing resistance to Bt cotton in

Telangana, Andhra Pradesh?

Farmers in the traditional cotton belt of Punjab are increasingly switching

to horticulture or paddy. As per the state agriculture department, the
cotton cultivation area has decreased from 421,000 hectares in 2014-15
to 248,900 hectares in 2022-23.
The production of cotton has halved from 1,347 bales to 444 bales in the
period. Whereas paddy cultivation has increased from 2,895,000 hectares
to 3,167,800 hectares for the same timeframe.

Rajvinder Singh (72), from Teona Pujarian village in Bathinda, Punjab, said
most cotton farmers in the region have switched to paddy cultivation.

“The soil and the water pH levels are not suitable for paddy here. I
experimented with eight acres of paddy, out of which I lost three. But the
losses are still better than cotton, which results in losing the entire
investment,” he said.

Many farmers switching to paddy are not thinking about the long-term
impacts as paddy has a very high water requirement, an agriculture
officer said on condition of anonymity. “Exploiting high levels of
groundwater will have environmental repercussions. But immediate profits
and economic gains become more crucial for survival than future
implications,” he said.

Read more: Study points to quality issues in Bt cotton seeds

Farmers in Rajasthan do not even have the option of diversifying their

crops. “The soil is unfit for paddy cultivation, and the water is saline. We
are stuck with growing cotton even if it runs into losses,” said Ram Pratap,
a resident of Bamboowali Dhani in Hanumangarh, Rajasthan.

Pratap said there was no other alternative for farmers like him. “Our
geographic location forces us to stick with cotton farming and the
government must help us out by bringing in improved seeds to the
market,” he said.

“Pink bollworm attacking Bt Cotton crops was first identified in 2008, but
the authorities are still not coming up with seeds resistant to its attacks.
Why are we still using seeds created in 2006 and technology has not been
used to upgrade it,” Pratap asked.
 https://actascientific.com/ASAG/pdf/ASAG-05-0954.pdf

Management of Pink Bollworm Pectinophora gossypiella

(Saunders) Using Mating Disruption Pheromone (PB Rope L) in
Cotton
Roop Singh Meena1*, Pardeep Kumar1, Bharat Lal Meena2 and Bhagwan Singh Meena1

1 Agricultural Research Station, Sriganganagar, Sawmi Keshwananad Rajasthan Agricultural

University, Bikaner, Rajasthan, India 2 Krishi Vigyan Kendra, Sawai Madhopur, Agriculture University,
Kota, Rajasthan, India *Corresponding Author: Roop Singh Meena, Agricultural Research Station,
Sriganganagar, Sawmi Keshwananad Rajasthan Agricultural University, Bikaner, Rajasthan, India.
2021

Abstract
The efficacy of mating disruption pheromone (PB Rope L) was evaluated against pink bollworm
Pectinophora gossypiella (Saunders) to monitoring pheromone traps catches as well as by sampling
of bolls to record larval population and damage levels in cotton fields during Kharif 2016 and 2017, in
north zone of India. This experiment was conducted in large plot in 10.0 ha and 4.0 ha area reserved
as control. The main field T1 and T2 divided into two subplots T1 A and T1 B, T2 C and T2 D of equal
size. In each of the subplots, six quadrates approximately 10x10 meter in the center at equal
distance were marked. The treated fields were compared with control fields, three to four
insecticides sprays in T1 A and T1 B, whereas 6 spray in T2 C and 9 sprays were applied in T2 D. The
dispensers reduced pink bollworm catches in pheromone trap and reduced larval population and
crop damage. Results showed that moth catches were highly suppressed in the treated fields by
installation of the dispensers @ 685 PB Rope L at 50 days after crop sowing. Damage percentage was
significantly reduced in treated fields as compared to control fields; Significant difference was
observed in average seed cotton yield/ha. This study recommends the use of single application of
pheromone dispenser at 50 days after crop sowing. These Rope were sufficient to maintain the moth
catches and percent green boll infestation at the lowest level as compared to untreated field.

Keywords: Cotton; PB Rope L; Pink Bollworm; Mating Disruption

 http://researchjournal.co.in/online/AJBS/AJBS%201(2)/1_A-68-69.pdf

Management of pinkbollworm, Pectinophora gossypiella (Saunders)

with PB ropel and IPM approach
Radhika P.*1 and Sahadeva Reddy B.2

1 Department of Entomology, Regional Agricultural Research Station, Nandyal 518 503 (A.P.) India 2
Department of Agronomy, Regional Agricultural Research Station, Nandyal 518 503 (A.P.) India
(Accepted : March, 2006)

Pheromone monitoring indicated advancement of pink bollworm activity to as early as August during
2004-05. Soon after emergence, the PBW larvae enter the fruiting body. As a result, farmers remain
totally ignorant about the damage caused by PBW and do not exercise any target specific control
measures about it . Considering the emerging status of this pest investigations were undertaken to
find out suitable measures to contain PBW, a large scale demonstration was undertaken in the
farmers fields to evaluate the suitability of PB Rope L. Moth catches in the control plot were
comparatively very high(12,478) throughout the crop season as compared to PB Rope L treated
plots(198). Thus the PB Rope L treated plot recorded the lowest percent boll and locule damage
compared to untreated plot and recorded highest yield and highest per cent good kapas. Key words :
Pinkbollworm, IPM, PB ropel, Control.
 https://wcrc.confex.com/wcrc/2007/techprogram/P1712.HTM

Tuesday, September 11, 2007

Mating disruption using PB Rope L: A promising option for pink

bollworm (Pectiphora gossypiella Saunders) management in cotton
Dr. S.B. Patil1, Dr. O.M. Bambawale2, Dr. R.K. Tanwar2, Dr. S.S. Udikeri3, and Ms. Renuka H.B.3. (1) Agricultural
Research Station, University of Agricultural Sciences, Dharwad Farm, DHARWAD 580 007, India, (2) National
Institute of Intigrated Pest Management, IARI Capus, PUSA, NEW DELHI, India, (3) Agricultural Research
Station, University of Agricultural Sciences, Dharwad-580 007, Dharwad Farm, DHARWAD, India

Integrated management of pink bollworm through mating disruption technique using a sex pheromone
formulation offers a practical and ideal approach. Large scale field experiments were conducted during 2005-06
and 2006-07 at agril research station Dharwad to evaluate the efficacy of PB Rope L (Sex pheromone based
commercial product) for management of pink bollworm. PB Rope L dispensers @ 200/ha were tied to cotton
stalks at pinsquare stage. The performance was compared with the RPP. The mating disruption block registered
significantly lower moth trap catches. The extent of reduction was 79.38 and 93.81 per cent over RPP during the
consecutive years. The average reduction in green boll damage was 72.62 and 70.11 per cent. The reduction in
the larval population in fruiting bodies was 52.34 and 56.70 per cent over RPP. Significant difference with respect
to open boll was noticed. The extent of reduction was 68.86 and 66.17 per cent over RPP. The reduction in
locule damage was quite conspicuous where mating disruption was exercised. Significantly higher seed cotton
yield (32.12 and 40.92 %) was recorded over control (RPP).

Pink bollworm Pectinophora gossypiella (Saunders) is one of the key pests of cotton, emerging as a real threat
to cotton cultivation in southern and central parts of India inflicting locule damage to an extent of 55 per cent and
reducing the seed cotton yield to an extent of 35 to 90 per cent (Narayanan 1962) ascertained that 75 to 100 per
cent bolls are liable to be damaged by this pest. Agarwal and Katiyar (1979) calculated yield loss to an extent of
6525 MT annually. Its infestation causes premature opening of bolls resulting in stained immature fiber
(Agarwal et al., 1984).Significant reduction in fiber properties of lint from infested bolls has been observed by
Shiva Subramanian, 1991.

The severity of pink bollworm infestation causes both qualitative and quantitative loss heavily and it is much more
pronounced in interspecific hybrids which are important sources of extra long staple cotton and is of utmost
importance from the point of view of textile industry and export. Pink bollworm endemic nature and reproductive
capacity seriously challenged all control efforts. Besides this pest is inaccessible to routine methods of pesticide
control and even conventional IPM components do not provide acceptable level of management. As a possible
means the use of certain behavior modifying chemicals are potentially a viable alterative to the use of
conventional insecticides and reached the stage of commercial production. Sex pheromone formulation of pink
bollworm containing ZZ/ZE-7-11 Hexa decadien-1-yl-Acetate as active ingredient has been commercially used
world over as an integrated component of IPM for successful management of PBW incidence in particular and
other insect pest in general in cotton. Insect pest management through mating disruption technique using sex
pheromone offers a practical and ideal approach to combat the above complex situation prevailing in cotton
ecosystem (Patil.et al.,2004). Considering the severity of this dreaded pest, effort has been made to find out a
suitable control measure a large scale field experiment was undertaken at ARS, Dharwad farm to evaluate the
efficacy of PB Rope L [Sex -pheromone formulation of Pink bollworm manufactured by Shin-Etsu Chemical Co.
Ltd., Japan. supplied by New chemi industries Ltd., Mumbai, India during 2005-06 and 2006-07 cropping periods.

Materials and Methods

Large scale field trials were conducted during 2005-06 and 2006-07 in farmers’ field to evaluate the suitability of
PB Rope L for the management of pink bollworm. To compare PB Rope L in conjunction with insecticide
schedule for the control of pink bollworm on cotton, field experiments were conducted at Agricultural Research
Station, Dharwad farm during Kharif 2005-06 and 2006-07 on an area of 20 hectares. The popular inter specific
cotton hybrids viz., RAHB-87, DHH-11, Sahana and Bt genotypes on an a area of 20 hectare were growth.
However, observations were recorded only in RAHB-87 block which was compared with farmers field at
Yammatti (Tq: Kalagatagi) where RAHB -87 was grown on an area of five acres. The treatments followed in the
experiments were as detailed below.

Block A- PB Rope L @ 200 / ha + ETL based/chemical application (Table-1)

Block B- Control (No PB Rope L) farmer’s schedule of spray only (Table-1)

PB Rope L dispensers containing synthetic sex pheromone of pink bollworm [ZZ/ZE-7,11-Hexadecadiene-1-yl-

Acetate] were tied manually by twisting the dispenser around central shoot of the cotton plant at first pin square
stage (40- 45 days after sowing). The performance of PB Rope L in combination with plant protection schedule in
reducing the pink bollworm incidence was compared with recommended plant protection where there was no
target specific treatment for pink bollworm management. These fields were located 2-3 km away from
experimental area.

Plot A: PB Rope L dispensers @ 200 / ha were tied and pesticide application was made strictly based on
economic threshold level of pink bollworm moth trap catches.

Plot B: farmers applied pesticides as per their choice and agronomical practices and cotton hybrid grown were
similar to that of Plot A.

A common spray of Acetamiprid 20 SP @ 0.2 g / l was applied to protect the crop from sucking pest complex at
25th and 40th days after sowing. Similarly to protect the crop from bollworms (Earias spp.
and Helicoverpa armigera) recommended bollworm management sprays were initiated from 45 days after sowing
and continued at 10 to 12 days interval till the bollworms activity ceased.

Results and discussion

The bio efficacy of PB Rope L in reducing the PBW incidence was evaluated by recording pheromone trap
catches, green boll infestation, larval population, open boll damage, locule damage at harvest and by comparing
the seed cotton yield.

Observations on moth trap catches were initiated at weekly interval starting from first week of September and
confined up to January during both the cropping period. The density of male moths varied to a greater extent
between both the blocks with the advancement of cropping period. Significantly lower moth trap catch was
observed in block ‘A’ where target specific treatment PB Rope L was applied in addition to regular plant
protection schedule. The number of moths trapped in block ‘A’ ranged from 1.80 to 18.40 with a mean of
5.92/trap/night as compared to 7.00 to 110/trap/night with mean of 28.72 moths and 0.07 to 3.63 moths with a
mean of 1.03 as against 4.26 to 66.82 with a mean of 21.69 moths/trap/night during 2005-06 and 2006-07
respectively. The extent of reduction in moth trap catches was 79.38 and 93.81 per cent during the respective
years. Mean trap catches during the consecutive years was 24.69 and 3.41 with a reduction of 86.18 per cent
(Table-2). The efficacy of PB Rope L in suppressing the activity of pink bollworm was quite convincing in terms of
moth trap catches over the season. These findings are in agreement with Patil et al., (2004) who recorded an
average of 0.8 moths/ trap/ night when PB Rope L was used @ 200 dispensers/ ha as against 2.6 / trap / night in
control plots.

Significant difference with respect to green boll infestation was observed between the blocks (Table-3). Initially
the numbers of green bolls damaged were found to be less. As the season advanced the green boll damage
increased reaching peak during January moth.

The mean percentage of green boll damage in block ‘A’ was 8.00 and 7.50 with reduction of 72.62 and 70.11 per
cent during 2005-06 and 2006-07 respectively. Over the season, the extent of reduction in green boll damage
was 72.04 per cent over RPP alone where there was no target specific treatment for PBW management.
Reduced incidence on green boll was the reflection of differential larval population that prevailed in blocks with
and without PB Rope L. Pink bollworm made its first appearance during September month during the
consecutives seasons and maximum number of larvae were observed during January month. Significantly, lower
population to larvae was recorded in treated block (9.76 and 8.40/ 50 bolls) as compared to control block where
ETL based plant protection alone was taken (20.48 and 19.40/50 bolls). The extent of reduction in larval
population was 52.34 and 56.70 per cent over the control block during 2006-05 and 2006-07 respectively. Over
the season (Table 4). The mean larval population in PB Rope L treated block was 8.75/50 bolls as against 19.60
in control block with a reduction of 55.35 per cent Collor et al., (2004) reported effective control of pink bollworm
with controlled release of formulation of sex pheromone of PBW mating disruption technique. These findings are
also in close agreement with the findings of Qureshi et al. (1988) who observed a significantly low incidence of
pink boll worm larvae in green bolls (1.67%) while using PB Rope L @ 55 g ai / ha. It was interesting to note that
damage to open boll was also lowered by the treatment containing PB Rope L. As against 47.83 and 44.02 per
cent open boll damage in control block, 14.89 and 14.89 per cent was recorded in treated blocks during the
respective years (Table -5). The mean PBW incidence on locule basis was significantly low i e., 11.55 and 11.81
per cent in PB Rope L block as compared to 48.56 and 47.97 per cent in control block during the consecutive
seasons. The extent of reduction was 76.21 and 75.38 per cent respectively with a mean of 75.75 per cent (Table
5). These findings are in confirmly with the observation of Sohi et al., (1999) who recorded significantly low per
cent open boll and locule damage in the fields treated with mating disruption along with the insecticide compared
to field treated with insecticide alone who reported less PBW infestation in green bolls and locules in PB Rope L
treated block.

The present studies conducted during 2005-06 and 2006-07, for the management of pink bollworm using PB
Rope L @ 200 dispensers/ha applied at pin square stage of cotton proved to be promising option in endemically
infested areas. It is claimed that PB Rope L when tied to plants, natural instinctive flight behaviour of male moths
bring the adults in more prolonged contact with treated plants. Depletion of male moths in the environment
results in reduction of laying fertile eggs and thus brings down larval infestation and crop damage. Significantly
higher yield of 22.95 q/ha was harvested from PB Rope L treated block accounting for 32.12 per cent increase
over RPP block (17.37 q/ha). Further during 2006-07, the seed cotton yield of 23.52 q/ha with an increase of
40.92 per cent was recorded in PB Rope L treated block. Pooled data revealed that significantly higher (23.24
q/ha) seed cotton yield was harvested in block as against 17.03 q/ha from RPP block with 36.46 per cent
increase (Table 6). Net return from PB Rope L treated block was Rs.30773 and Rs.31273 as against Rs.220013
and Rs.21353 in RPP during 2005-06 and 2006-07 respectively (Table 7). The field study conducted at Raichur,
Karnataka also revealed that the application of PB Rope L along with insecticidal spray recorded lower open boll
and locule damage and higher seed cotton yield (Patil et al., 2004). In the present investigation, there is a
marked increase in yield as well as net returns. Considering the net returns and yield advantage, application of
PB Rope L in addition to the regular plant protection schedule is found to be economically viable and promising
option for pink bollworm management.

References:

AGARWAL, P.A., GUPTA, A.P AND GARG, D.O., 1984, Cotton Pest Management. Research Co-publications,
East Azad Nagar, Delhi, p.941.

AGARWAL, R.A., AND KATIYAR., 1979, An estimate of losses of seed kapas and seed due to bollworms on
cotton in India. Indian Journal of Entomology.41 (4) : 143-148.

COLLOR U.R., QUIJO, J. L AND GALVEZ MANZANO V 2003 , Mating disruption strategies for the control of
pink bollworm in cotton fields. Boletin de Sanidad vegetal, Madrid, Sapin. Pages 28 (7), 193-198.

GRANT, G.G., 1978, Field trials on disruption of pheromone communication of Tussock moths. Journal of
Economic Entomology, 71: 453-457.

NARAYANAN, E.S. 1962, Biology and Method of Control of Some Important Insect Pests of Cotton in
India. Indian central Cotton Committee Publication, Bombay, p.44

PATIL, B.V., BHEEMANNA, M., HANCHINAL, S.G. AND ANANDKUMAR, V. 2004, Management of Cotton Pink
bollworm using PB Rope L mating disruptant. In The proceedings of International Symposium on “Strategies for
Sustainable Cotton Production- A Global Vision” held at University of Agricultural Sciences, Dharwad, Karnataka
(India) from 23-25 November 2004. pp:172-175.

QURESHI, Z.A, M.D. ARIF, N. AHMED AND NAZEEBULLAH 1988, control of pink bollworm by mating disruption
technique. Pakistan. J.Sci.Ind. Res., 31, 711-713.

SHIVASUBRAMANIAN, P., 1991, Ecology host plant insect interactions as management of pink boll worm, y
Pectinophora gossypiella (Saunders) on cotton .Ph.D. Thesis, Tamil Nadu Agricultural University, Coimbatore.

SOHI, A.S., JAGINDER SINGH AND RUSSEL, D., 1999, Further studies of mating disruption in pink
bollworm P.gossypiella (Saunders) using sex pheromone as a component of IPM programme in irrigated cotton
fields in Punjab. Pest Management and Economic Zoology, 7 (1): 31-38.

Table 1: Plant protection schedule followed in both experimental blocks

Block A Block B

(RPP + PB Rope L) (RPP alone)

Sl. No. Chemical sprayed Dosage Chemical sprayed Dosage.

 1. Endosulfan 35EC 875 g/ha Endosulfan 35EC 875 g/ha

2 Quinalphos 25 EC 500 g/ha Quinalphos 25 EC 500 g/ha

3. Acephate 75 SP 700 g/ha Acephate 75 SP 700 g/ha

4. Profenophos 50 EC 500g/ha Profenophos 50 EC 500g/ha

5. λ-Cyhalothirn 5 EC 25g/ha λ-Cyhalothirn 5 EC 25g/ha

6. Indoxacarb 15 SC 0.5 L/ha Indoxacarb 15 SC 0.5 L/ha

7 PB rope L 200 dispensers /ha -- --

Common spray schedule followed for sucking pest management.

1. Spraying of Imidacloprid 200 SL @ 100ml/ha

2. Spraying of Acetamiprid 20 SP @ 50 g/ha

Table 2: Pheromone trap catches of Pink boll worm moths

Artificial Intelligence In Indian Agriculture
Author: Anmol M. Dua
Consultant, Bhajan Global lmpact Foundation

Tag(S) : Agriculture & Rural Economy, Technology

Introduction

In India, the agriculture and the allied sector contributes less than 17% to its US$ 3
trillion economy today (Sunder, 2018). Average annual farmer incomes remain
abysmally low at ₹77,976 (or US$ 1,128; Financial Express, 2018). In terms of land
and labour inputs, nearly 50% of India’s land is tilled (Agriculture Census, 2015-16),
and agriculture directly employs 41.1% of India’s working age population (World
Bank, 2018).

This low output to input ratio points to the huge systemic flaws in India’s agronomy
that brings severe hardship to farmers and agricultural labourers, who bear the brunt
of rising input costs, falling productivity, climate change, water scarcity, poor market
access, technological stagnation, and so on. Although the agricultural predicament in
India needs to be addressed at multiple levels, this article discusses the role of
technology – in particular, artificial intelligence (AI) – in increasing agricultural
output, and therefore, farmer incomes in India.

AI in Agriculture – An Overview

Essentially, AI in agriculture is segmented into robotics, soil and crop management,

and animal husbandry, and is designed to make farming easier, more accurate, more
profitable, and more productive for the farmer. It is estimated that AI and connected
farm services can impact 70 million Indian farmers by 2020, thereby adding US$ 9
billion to farmer incomes. In 2017, the global AI in agriculture market size was US$
240 million, and is expected to reach US$ 1.1 billion by 2025 (Maher, 2018).
Furthermore, issues such as population growth, climate change, and food security
demand innovative ways to improve crop yield. Therefore, understanding the use of
AI in agriculture becomes compelling.

Applications of AI in Agriculture
The world needs to produce 50% more food by 2050. However, only 4% of additional
land will come under cultivation to meet this demand (Bagchee, 2018). AI holds the
promise of driving an agricultural revolution at a time when the world must produce
more food using fewer resources. This essay briefly discusses key applications of AI
in agriculture along different stages of the cropping cycle that have the potential to
pay dividends to farmers in efficiency gains and higher incomes.

 Soil Analysis and Monitoring

AI can be used to monitor soil health with the help of sensors, cameras, and infrared
rays that scan the soil for its nutritional properties (Sennaar, 2019; Baruah, 2018).
This also helps in understanding the reaction of specific seeds to different soils, the
impact of weather changes on the soil, and the probability of the spread of diseases
and pests (Irimia, 2016). With such data in hand, the efficiency of crop inputs is
improved, leading to cost savings and productivity gains for farmers. Currently, an
average of 207.56 kg of chemical fertilisers are used per hectare in Haryana
annually (one of the highest among Indian states). Besides being costly for farmers,
fertilisers also introduce harmful substances into the food chain through crops and
the water table (Indian Fertiliser Scenario, 2013).

Case Study: AI-backed soil health monitoring used in Raleigh, North Carolina, USA,
led to huge efficiency gains in the use of agro-inputs by cutting the use of chemical
fertilisers by nearly 40% (Sennaar, 2019). Furthermore, the spatial analysis
capabilities of geographic information system (GIS) technologies helps in efficient
water management during irrigation. For instance, in Alfalfa in Riverdale, California,
GIS technologies in irrigation helped increase the per acre crop output by up to
37.5%, and reduced water usage by 20% (Fictchett, 2013).

Therefore, AI in analysing and monitoring soil health helps to improve the

sustainability of a given piece of arable land.

 Crop Sowing

AI in crop sowing is used essentially to drive predictive analytics to determine when

and how to sow. It helps in making predictions on the right time to plant, apply
fertilisers, harvest, bale, till, etc. based on climate data, historical conditions, market
conditions for inputs and outputs, personal information, and so on. Crops can also be
sowed using AI-aided machinery at equidistant intervals and at optimal depths.

Case Study: In 2016, Microsoft, in partnership with ICRISAT (The International Crop
Research Institute for the Semi-Arid Tropics), initiated a pilot project in Devanakonda
Mandal in the Kurnool district of Andhra Pradesh. The pilot had a sample base of 175
farmers who were alerted on their mobile phones about suitable cropping dates, land
preparation, and soil test-based fertiliser utilisation. This helped increase crop output
by around 30%. In 2017, this project was expanded to cater to approximately 3,000
farmers in Karnataka and Andhra Pradesh during the Kharif cycle for a host of crops
like groundnut, ragi, maize, rice, and cotton, among others. The increase in crop
yield following the AI intervention ranged from 10-30% across all crops (Nagpal,
2017).

Therefore, AI in crop sowing has the potential to increase per acre crop
output as well as decrease input costs for farmers.

 Weed and Pest Control

Average losses of up to 90% of the total crop production have been reported due to
the infestation of weed (Meena, 2015). Similarly, average losses of up to 19% have
been reported due to pests (Dhaliwal et al., 2015). This leads to a greater use of
pesticides, further contaminating the soil and groundwater.

Table 1: Effects of Weeds on Crop Yields in India

 (Source: Meena, 2015)

As of today, there are 250 identified species of weeds which have become
completely resistant to herbicides (Sennaar, 2018), presenting a severe threat to the
sustainability of crop production. Pesticide resistance is also on the rise. The
purchase of insecticides and pesticides contribute approximately 5% to the total cost
inputs in agriculture, and this cost is on rise both in percentage and absolute terms
(Price Policy for Kharif crops, 2017-18; Price Policy for Rabi Crops, 2014-15).

Case Study: AI helps to optimise on weed and pest management by reducing up to

80% of weedicides and pesticides used currently (Sennaar, 2019). According to a
Bloomberg report (2018), a project was undertaken by the Blue River Technologies
(now under John Deere & Co.) to evaluate the effectiveness of AI-supported weed
controlling technologies. The ‘See & Spray’ technology tested under this project in
Marianna, Arkansas, USA, reduced the expenditure on weedicides by around 90%
reducing the need for weedicides used per acre from 20 to just 2 gallons (Little,
2018).

Therefore, AI finds great application in precision weed and pest

management.

 Crop Harvesting

An estimated 40% of annual agriculture costs go into the employment of labour,

predominantly for sowing and harvesting (Sennaar, 2019). AI-enabled robots for
harvesting can lead to huge cost savings by reducing the need for approximately 4
agricultural labourers per acre of land (Panpatte, 2018). Furthermore, crops can be
sorted according to pre-identified grades at the time of harvest, saving time and
enhancing the quality of crops. However, AI is likely to change the way labour is
employed in agriculture. Although conventional manual jobs will be replaced, AI
presents new opportunities for job creation.

Case Study: AI-enabled robots are being widely deployed on tomato farms in Japan,
and have reduced the on-field labour time by 20% (Panasonic, 2018).

Therefore, AI-enabled robots for crop harvesting save the resources of

labour and time for farmers.

 Supply Chain Management

Policymakers have not yet been able to tackle the agricultural supply chain
challenge. On the one hand, farmers either do not receive a suitable price for their
produce that continues to rot in mandis (or marketplace), and on the other, food
consumers either end up paying exorbitant prices or are malnourished. Although AI
in agricultural supply chain management is yet to make major inroads, its informed
application in supply chain planning and optimisation, including demand forecasting
and logistics, can lead to huge cost savings for farmers, and solve the information
asymmetry problem for buyers.

Case Study: Platforms like Jivabhumi’s ‘Foodprint’ aggregates and traces farm
produce, and offers a digital marketplace. It uses blockchain technology to connect
farmers and institutional buyers via an efficient and traceable supply chain (Baruah,
2019).
Therefore, AI has the potential to improve the agricultural supply chain
efficiency and reduce spoilage.

Concluding Remarks

Unquestionably, the agricultural sector needs a greater impetus from policymakers

in addressing the aforementioned challenges. This essay suggests AI as a possible
way forward in increasing the economic and environmental sustainability of
agriculture. Certainly, AI is not without its own shortcomings. For instance, temporal
data collection challenges exist. As of today, AI in Indian agriculture is yet to
penetrate to the small and marginal farmer. AI runs the risk of further increasing
economic inequalities between large landholders and small landholders and landless
labourers. Therefore, any policy measure on this front needs to be carefully designed
and implemented. It is important for the private and public sectors to collaborate in
making AI and other technologies in agriculture affordable and properly understood
by the end users – the farmer.

References

Agriculture Census (2015-16). ‘All India Report on Number and Area of Operational
Holdings.’ : Agriculture Census Division, Department of Agriculture, Co-operation and
Social Welfare, Ministry of Agriculture and Farmers Welfare, Government of India, Pg.
6. http://agcensus.nic.in/document/agcen1516/T1_ac_2015_16.pdf

Bagchi, Arka (2018). ‘Artificial Intelligence in Agriculture.’

https://www.mindtree.com/sites/default/files/2018-04/Artificial%20Intelligence%20in
%20Agriculture.pdf

Baruah, Ayushman (2018). ‘Artificial Intelligence in Indian Agriculture – An Indian

Industry and Startup Review’. www.emerj.com https://emerj.com/ai-sector-
overviews/artificial-intelligence-in-indian-agriculture-an-industry-and-startup-
overview/

Dhaliwal, G.S. & Jindal, Vikas & Mohindru, Bharathi. (2015). Crop Losses due to insect
pests: Global and Indian Scenario. Indian Journal of Entomology.

Employment in Agriculture (2018). The World Bank Report. www.worldbank.org

https://data.worldbank.org/indicator/SL.AGR.EMPL.ZS
FE Bureau (2018). ‘Economic Survey 2017-18: Agriculture- Climate change likely to
lower farmer’s income by 25%.’. Financial
Express. https://www.financialexpress.com/budget/economic-survey-2017-18-
agriculture-climate-change-likely-to-lower-farmers-income-by-25/1035560/

Fictchett, Tom (2013). ‘Netafim Drip Irrigation Success Story’ Western Farm
Press. https://www.netafimusa.com/wp-content/uploads/2016/08/Alfalfa-Success-
Maddox-2013.pdf

Geography Statistics of
India; www.worldatlas.com https://www.worldatlas.com/webimage/countrys/asia/
india/inlandst.htm

Hunger Facts (2019). www.bhook.com http://www.bhookh.com/hunger_facts.php

Indian Fertiliser Scenario (2013). Department of Fertilisers, Ministry of Chemicals and

Fertilisers, Government of India.

http://fert.nic.in/sites/default/files/Indian%20Fertilizer%20SCENARIO-2014.pdf

Irimia, Madalina (2016). ‘Five ways agriculture could benefit from artificial
intelligence.’ AI for the Enterprise, IBM.

https://www.ibm.com/blogs/watson/2016/12/five-ways-agriculture-benefit-artificial-
intelligence/

Maher, Tooba (2018), ‘A Complete Overview of Artificial Intelligence in Agriculture

Market.’ https://krishijagran.com/news/a-complete-overview-of-artificial-intelligence-
in-agriculture-market/

Meena, Ram Swaroop. (2015). ‘A Review of Weed Management in India: The Need of
New Directions for Sustainable Agriculture’. 10. 253-263.

Nagpal, Jasmeen (2017). ‘Digital Agriculture: Farmers in India are using AI to

increase crop yields.’ Microsoft India News Center.
www.microsoft.com https://news.microsoft.com/en-in/features/ai-agriculture-icrisat-
upl-india/

Little, Amanda (2018). ‘This Army of AI Robots will Feed the World.’ Bloomberg
Businessweek. https://www.bloomberg.com/news/features/2018-01-11/this-army-of-
ai-robots-will-feed-the-world

Panasonic (2018). ‘Introducing AI-backed Tomato Harvesting Robots’.

https://news.panasonic.com/global/stories/2018/57801.html
Panpatte, Deepak. (2018). ‘Artificial Intelligence in Agriculture: An Emerging Era of
Research.’

‘Price Policy for Kharif Crops’ (2017-18). Commission for Agricultural Costs and
Prices, Department of Agriculture, Cooperation and Farmers Welfare, Ministry of
Agriculture and Farmers Welfare, Government of India.

‘Price Policy for Rabi Crops’ (2014-15). Commission for Agriculture Costs and Prices,
Department of Agriculture, Cooperation and Farmers Welfare, Ministry of Agriculture
and Farmers Welfare, Government of India.

Sennaar, Kumba. (2018). ‘AI in Agriculture – Present Applications and

Impact.’ www.emerj.com https://emerj.com/ai-sector-overviews/ai-agriculture-
present-applications-impact/

Stoltzfus, Justin (2018). ‘The 6 Most Amazing AI Advances in

Agriculture.’ www.techopedia.com

https://www.techopedia.com/the-6-most-amazing-ai-advances-in-agriculture/
2/33177Sunder, Shushrut (2018). ‘India economic survey 2018: Farmers gain as
agriculture mechanisation gains up, but more R & D needed.’
https://www.financialexpress.com/budget/india-economic-survey-2018-for-farmers-
agriculture-gdp-msp/1034266/
https://www.nature.com/articles/s41598-020-80184-6

Degree day-based model predicts pink bollworm

phenology across geographical locations of
subtropics and semi-arid tropics of India. Sci Rep 11,
436 (2021).

Fand, B.B., Nagrare, V.S., Bal, S.K. et al.

 Babasaheb B. Fand,  V. S. Nagrare,  S. K. Bal,  V. Chinna Babu Naik,  B. V. Naikwadi,  D. J.

Mahule,  Nandini Gokte-Narkhedkar &  V. N. Waghmare

ABSTRACT

There is a global concern about the effects of climate change driven shifts
in species phenology on crop pests. Using geographically and temporally
extensive data set of moth trap catches and temperatures across the
cotton growing states of India, we predicted the phenology of cotton pink
bollworm Pectinophora gossypiella (Saunders). Our approach was
centered on growing degree days (GDD), a measure of thermal
accumulation that provides a mechanistic link between climate change
and species’ phenology. The phenology change was predicted by
calculating absolute error associated with DD and ordinal date, an
alternative predictor of phenology, for peak moth abundance. Our results
show that GDD outperformed the ordinal dates in predicting peak moth
abundance in 6 out of 10 selected locations. Using established thresholds
of 13.0/34.0 °C, mean DD accumulated between the consecutive moth
peaks across different years were estimated at 504.05 ± 4.84. Seven
generations were determined for pink bollworm in a cropping season, the
length of which varied between 35 and 73 days in response to
temperature. Pink bollworm population reached its peak during third
generation which can be the target for management actions. The study
provides essential information for developing pink bollworm management
strategies under climate change

https://krishi.icar.gov.in/jspui/bitstream/123456789/7100/1/Approaches%20to
%20weather%20based%20prediction%20of%20insects%20a%20case%20study
%20on%20cotton%20pink%20bollworm.pdf

Approaches to weather based prediction of insects: a

case study on cotton pink bollworm Pectinophora
gossypiella. January 2011,
ind jr plantProt 39(3): 163-169. S. Vennila, M Agarwal, DSingh, P Pal, VK Biradar.

PBW severity could be predicted based on on mx temp > 34 deg C in 40th std week,
min temp < 17 deg C in 44th std week, relative humidity < 33 & 70% resp in evening
& morning in 4 &46th std week, dry spell followed by rains in 41th & 42nd std week.
Validation using data during 2001 to 2004 provided over & under estimate extent.
https://www.researchtrend.net/bfij/pdf/122%20Monitoring%20of%20Pink
%20Bollworm,%20Pectinophora%20gossypiella%20(Saund.)%20throughout
%20the%20Cropping%20Season%20by%20Gossyplure%20Lalsingh
%20Rathod.pdf
Biological Forum – An International Journal 14(3): 728-731(2022)

Monitoring of Pink Bollworm, Pectinophora gossypiella (Saund.)

throughout the Cropping Season by Gossyplure
Lalsingh Rathod* , A.V. Kolhe, D.B. Undirwade, A.K. Sadawarte, S.K. Bhalkare and P.K. Rathod
Department of Entomology, Dr. Panjabrao Deshmukh Krishi Vidyapeeth, Akola, (Maharastra, India)
(Corresponding author: Lalsingh Rathod*) (Received 09 June 2022, Accepted 26 July, 2022)
(Published by Research Trend, Website: www.researchtrend.net)
ABSTRACT:
Pink bollworm is the most serious and destructive cotton pest, causing locule damage to the tune of
55% and a reduction in seed cotton yield ranging from 35% to 90%. (Narayanan, 1962). This pest
must be monitored on a regular basis in order to gain insight into seasonal population fluctuations
and achieve effective control. Pesticides are widely used as a major weapon in the fight against this
pest damage. In India, total insecticides used to control bollworms alone were 9410 MT in 2003-04,
valued at 747.6 crores (Kranthi, 2012). Despite consuming such a high percentage of pesticides, pink
bollworm pressure could not be controlled and instead worsened the situation by becoming
resistant to both Bt toxins and major insecticides. It is now necessary to switch from conventional
management methods to cutting-edge, economical, and environmentally friendly pheromone-based
techniques to combat PBW in Bt cotton. Considering all those things in view an experiment was
undertaken at the cotton research unit, Dr. Panjabrao Deshmukh Krishi Vidyapeeth, Akola,
(Maharashtra) for two years (2019-20 and 2020-21). During the year 2019-20 a sudden increase in
the number of moth trap catches was witnessed during December’s second fortnight (corresponding
to 52nd SMW / first cotton picking) i.e. 341 moth/ trap/ fortnight and continued at a high level till
the end of February’s first fortnight (corresponding to 7th SMW/second and third cotton picking).
Thereafter the moth trap catches showed a gradual decrease in number in subsequent months.
Similarly, the pink bollworm moth trap catches recorded during the second (2020-21) year were
found marginally greater than in the first year and began earlier. The number of PBW moth trap
catches increased gradually reaching its peak during the December’s first fortnight (corresponding to
50th SMW/peak boll bursting) i.e. 376.4 moth/ trap/ fortnight and thereafter gradually declined.
Keywords: Pink bollworm, Pectinophora gossypiella, gossyplure, monitoring, pheromone trap
https://cicr.org.in/wp-content/uploads/CI-2022-22.pdf

Cotton News and Innovations – February 2022 Research News Item Mass trapping - An Eco-friendly
tool for the Management of Pink Bollworm in Cotton
V. Chinna Babu Naik, B.B. Fand and Y.G. Prasad ICAR-Central Institute for Cotton Research, Nagpur-
440 010, India, E-mail: chinnaenton@gmail.com
Cotton is a major fibre crop cultivated across the country. The major factor limiting the yield is insect
pest damage, in specific, the pink boll worm. Bt cotton with transgenic technology has been
introduced in India to combat bollworms of cotton. But the development of resistance in bollworms
diminishes the efficacy of Bt technology. Pink bollworm causes significant economic yield loss to the
crop. Moreover, the larvae of pink bollworm usually lie within the cotton bolls making them
unreachable to insecticidal sprays due to which, its management has become a difficult task.
Moreover, the use of insecticides has resulted in the problem of resistance, resurgence of pests and
unwanted pesticide residues in the environment. Hence an eco-friendly tool for bollworm
management in view of increasing menace of pink bollworm is presented herewith. Formulation of
pink bollworm pheromone lures, developed by ICAR-CICR was used for experimental trial in mass
trapping. Normally pheromone traps are used to detect both the presence and density of the pest
species. The important components of IPM include estimation of insect population, detection of new
areas of infestation at a very early stage and monitoring and timing of the insecticidal applications.
Earlier, the traps were used for mass trapping @ 8 traps /acre. Field experiments were conducted by
ICAR-CICR Nagpur for the evaluation of the mass trapping in managing pink bollworm using traps at
different densities viz., 8, 20, 25, 30, 35 and 40 per acre in research and farmers’ fields. To gain the
full benefit of mass trapping, all traps should be in the field by the recommended time from 45 days
after sowing i.e., two weeks prior to first flower to reduce the establishment of first-generation
larvae in the flowers. The field trial was conducted on 400 acres of farmers’ fields at Adilabad,
Gadchirolli, Chandrapur, Amravati and Wardha districts during 2021-2022. Nearly 6,000 pheromone
traps and 18,000 lures were placed in these fields for mass trapping of adult moth’s pink bollworm.
The trap densities of 30 and 35 traps per acre were found optimum in reducing the pink bollworm
damage effectively. The infestation of Pink bollworm was reduced by 10-20 % in mass trapped fields
as compared to control plots. There was reduction of one-two sprays with yield benefit observed
between 1-1.5 q seed cotton /acre. Impact of the project The acceptance of technology is the key
factor in adoption. In order to accept and adopt, the technology should be cheap, efficient and easy
to apply in the field. All these apply well to the mass trapping technology and thus can be adopted
by the farmers. The adoption of the technology has facilitated the farmers in reaping the additional
yield of nearly 1-1.5 quintal per acre. Apart from the monetary benefits, the additional advantages in
terms of safety to environment and natural enemies were also noticed while conducting the
experiment.
https://www.sciencedirect.com/science/article/abs/pii/S0261219423002478

Pink bollworm, Pectinophora gossypiella (Saunders) management

strategy, dissemination and impact assessment in India

Crop Protection
Volume 174, December 2023, 106424

Abstract
The recent emergence of pink bollworm (PBW), Pectinophora
gossypiella (Saunders) has posed serious ecological and economic
implications for cotton production in India since 2017. The project
“Insecticide Resistance Management (IRM): Dissemination of pink
bollworm management strategies in Bt-cotton” was implemented
consecutively for four years from 2018 -19, 2019–20, 2020-21 and 2021-
22 in 21 districts across three cotton growing zones of the country. In
each year, field demonstrations were conducted on 1050 acres of cotton
area and mass awareness was created through outreach activities. With
concerted and focused efforts, the PBW infestations could be brought
down by 42.75% in 2018–19, 43.49% in 2019–20, 35.55% in 2020-21 and
33.77% in 2021-22 in the demonstration plots over the farmer practices.
An increased benefit: cost ratios of 2.16:1, 2:1, 1.89:1 and 2.24:1 were
realized during 2018–19, 2019–20, 2020-21 and 2021–22, respectively.
Similarly, the average number of sprays for the control of cotton pests
during the season was reduced to 5.02, 4.94, 5.19 and 5.18 in integrated
pest management (IPM) fields as compared to 6.98, 7.66, 8.07 and 8.20
in non-integrated pest management (NIPM) fields in the corresponding
years. Reduction in pesticide usage in IPM vs NIPM fields was 46.39% in
terms of cost, and 38.96% in terms of volume during 2019-20 while
concurrent percent reduction in pesticide usage was 41.03 & 39.33
during 2020-21 and 40.70 & 39.38 in 2021–22. Comparatively more seed
cotton was harvested from IPM than NIPM fields i.e. 2190 vs 1820 kg/ha
in 2018–19, 2049 vs 1715 kg/ha in 2019–20, 1881 vs 1585 kg/ha in 2020-
21 and 1901 vs 1607 kg/ha in 2021–22. Significant success in reducing
PBW infestation, pesticide usage and corresponding improvement in
yield, economic benefits and environmental quality in terms of reduced
EIQFUR has been achieved through IPM demonstrations. This area-wide
success was achieved through collective efforts and active involvement of
all concerned stakeholders