论文信息
论文解读
AI 辅助整理- 方法
- 整合同行评议证据与公共部门数据,跨六类技术做多利益相关方综述。
- 发现
- 精准技术可增产15-30%、节水30-55%、降本10-20%,但小农获益不均。
- 机会点
- 缺乏长期田间试验、互操作标准与社会效益数据,小农适配与规模化落地是研究空白。
论文摘要
从投入要素到信息的转变是有意为之且意义重大的。本文结合经过同行评审的证据与公共部门数据,聚焦六个相互关联、正在从根本上改变农场内外决策方式的技术集群:基于人工智能(AI)的农场管理平台、物联网(IoT)智能灌溉、航空与卫星作物监测、区块链赋能的供应链数字化、可持续性与碳追踪工具,以及农场数据分析仪表盘。在所考察的文献中,精准工具与作物产量提升15%至30%、灌溉用水节约30%至55%、运营成本降低10%至20%相关联,而这些效果因作物种类、地理区域和农场规模而存在差异。本文一方面考察这些效应的作用机制,另一方面分析大型商业农场与小农户之间受益程度的差异,以及驱动实际采用的政策与基础设施条件。文章还指出了当前证据中存在的不足,例如缺乏长期田间试验、互操作性标准不完善以及社会成效数据缺失。文章最后讨论了如何将试点转化为粮食生产中持久的系统性变革,分别从政府与企业一侧以及中小生产者视角展开。
全文
Zenodo (CERN European Organization for Nuclear Research)Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 21
Real-World Impact: Transforming Agriculture with Technology A Multi-
Stakeholder Review of Smart Monitoring, Precision Irrigation, and Digital
Value-Chain Integration
Dr.A.Shaji George1, Dr.T.Baskar2, Dr.P.Balaji Srikaanth3
1Independent Researcher, Chennai, Tamil Nadu, India.
2Professor, Department of Physics, Shree Sathyam College of Engineering and Technology, Sankari Taluk,
Tamil Nadu, India.
3Associate Professor, Dept of Networking and Communications, School of Computing, SRM Institute of
Science and Technology, Kattankulathur, Tamil Nadu, India.
Abstract - There's a deliberate and significant transition from input to information . This article
combines peer -reviewed evidence and public -sector data on six interrelated technology clusters that
are fundamentally changing on-farm and off-farm decision making artificial intelligence (AI) based Farm
Management Platforms, Internet of Things (IoT) smart irrigation, aerial and satellite crop monitoring,
blockchain-enabled supply-chain digitisation, sustainability and carbon -tracking instruments, and farm
data analytics dashboards. In the literature examined, precision tools have been correlat ed with 15% to
30% gains in crop yields, 30% to 55% of water saved during irrigation, and 10% to 20% of operational costs
reduced, all with a range of variability based on the crop, the geographical region, and the size of the
farm. It examines the mechanisms of these effects on the one hand and the differential rates of benefit,
on the other, between large commercial farms and smallholders, and the policy and infrastructural
conditions that drive actual adoption. It also highlights areas of current evidence which are lacking, such
as limited long horizon field trials, poor interoperability standards, and lack of social outcome data. The
article ends with a discussion on how to turn pilots into lasting, syste m-level change in food production,
both from the government and enterprise side, as well as from the perspective of small and medium
producers.
Keywords: Precision agriculture, Internet of Things, Smart irrigation, Remote sensing, Blockchain
traceability, Carbon accounting, Agricultural policy.
-
INTRODUCTION
The agricultural sector experienced productivity increases mostly because of mechanization, use of
synthetic fertilizer, better cultivars, and extension of irrigation in most of the twentieth century. The input
driven model has tripled global cereal production between 1960 and 2000 but also had some well-known
side effects such as soil degradation, aquifer depletion, nutrient run-off, and low genetic diversity. In the
future, to continue increasing productivity, it will take a combination of different things. The size of arable
land per capita is shrinking, labour is becoming redundant in agriculture in most economies, and the
climate variability is increasing the productivity gap between potential and actual yields. In this context,
the spread of sensors, connectivity and analytical software has paved the way for a new type of
production decision, not based on convention, but on data, and at a sub-field or even per plant level.Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 22
This article looks at this transition. It focuses on the technologies that have progressed from
demonstration to field deployment within the past decade, and views them as a system, rather than
individual tools. The analysis is geared towards those with an interest in agronomy, policy, and digital
infrastructure, while at the same time remaining accessible to those with an interest in other fields, such
as public administration, development economics, or corporate sustainability. Giving a practical guiding
question. In cases where there is credible evidence of impact, what are the magnitudes, who is the
beneficiary, and what are the conditions that make a pilot project a lasting change in the way food is
produced and transported. It also takes seriously the boundaries of the current evidence base publication
bias for successful trials, lack of long term field evidence and potential for field evidence to be skewed by
these first movers size and funding.
-
OBJECTIVES OF THE STUDY
The study pursues four objectives. -
To describe and to characterise the current state of six technology clusters which are
transforming the state of agricultural practice and to provide a description that is technically
accurate but also accessible to non-specialists. -
To provide, in summary, the same impacts on yield, water use, operational cost and
environmental performance measured from peer -reviewed and public sector sources, with
confidence intervals and known limitations of the estimates. -
To assess the benefits across four groups of stakeholders national and sub-national governments,
society at large, commercial enterprises, and small and medium farm operators and the danger
that digital tools could exacerbate structural inequities. -
To define research and policy needs, and to suggest a research and policy agenda for scaling up
isolated deployments towards economically, environmentally, and socially justifiable system-level
transformation, in a sequenced fashion. -
HISTORICAL CONTEXT AND THE SHIFT TO INFORMATION-INTENSIVE FARMING
The smartphone isn't the first modern agricultural technology. From mechanical reaper to hybrid seed,
from Haber -Bosch process to chemical plant protection, from large scale irrigation engineering to the
long arc. All these waves boosted productivity significantly but had to be applied evenly in a non-uniform
field. The fundamental principle of precision agriculture, which was first expressed in the late 1980s with
the development of yield monitors to fit combine harvesters and the advent of the Global Positioning
System (GPS) for civilian use, is that variability within a field is great enough to warrant consideration of
subplots. Early commercial manifestations of that insight were early differential GPS equipment, variable
rate fertilizer spreaders, and grid soil sampling.
Defining economics of sensing radically shifted from 2010 onwards. Basic soil moisture probes have been
reduced to an affordable price that can be spread over a single season with a moderate -value crop.
Along with the cost of transmitting field data to the cloud, low -power wide-area networks like LoRaWAN
and NB -IoT reduced these costs. Multispectral payloads made their way into the consumer UAVs and
became affordable to agronomic consultants and large farms. In the early 2020s, adoption was
measured in official statistics in agencies in various countries. According to the United States DepartmentPartners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 23
of Agriculture Economic Research Service, in 2023, about one -fourth 27 per cent of farms in the United
States adopted at least one precision agriculture practice, with greater percentages of farms using yield
or soil mapping and auto-steering guidance on large scale row-crop farms 68 per cent and 70 per cent,
respectively compared to midsize crop farms 46 per cent and 52 per cent, respectively. In 2024, the size of
the US IoT market in precision agriculture was estimated to be around 2.36 billion dollars, w hile the global
market is expected to be around 32.75 billion dollars by 2027.
The present time is not just a continuation of this. There are three factors to indicate a qualitative change.
First, machine learning models are now sufficiently accurate to make useful inferences from the highly
variable and noisy field data with some machine learning models achieving 98 -99 per cent accuracy to
predict crop yield and recommendation under controlled laboratory conditions. Second, with the advent
of satellite constellations , high -resolution multispectral imagery can be acquired in days instead of
weeks, thus moving remote sensing from a retrospective to an operational tool. Third, food value chain
enterprises are under policy pressure to measure, rather than simply state, their environmental
performance, because of the climate reporting regime. With these three forces coming in combination,
digital agriculture has shifted from a productivity question to a strategic question.
Fig -1: Stylised adoption curve for selected precision agriculture practices on US farms, 2000 to 2023,
showing divergence between large -scale and small-scale operations. Data points are reproduced from
USDA Economic Research Service and USDA NASS Census of Agriculture .
-
CURRENT TRENDS IN AGRICULTURAL TECHNOLOGY
There are a few trends apparent in the field trials, market reports, and government data from 2024 and -
The first trend is the transformation of the value in agricultural technology stack from hardware to
software and data services. The price of sensors has been commoditized, and analytics, agronomic
decision support and traceability continue to be premium priced. Secondly, the restriction is
interoperability. Often, farms are using equipment from 3 to 4 different manufacturers, and the dataPartners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 24
formats don't match, reducing the value of the data asset and restrictive for analytics. Third, satellite
derived indicators and ground based sensors are integrating into hybrid architectures, where satellites
are used to achieve wide area coverage and ground sensors for calibration and depth. Fourthly, new
climate reporting frameworks, such as the European Corporate Sustainability Reporting Directive and
voluntary reporting standards compatible with the Task Force on Climate -related Financial Disclosures
are moving carbon accounting instruments from the fringes of agricultural technology to the centre.
The fifth, but little discussed, trend relates to financing. The greater number of precision tools sold are now
being packaged with insurance, credit, or offtake agreement. Parametric crop insurance platforms, based
on weather station or satellite indexes, and smart -contract based input financing have gone beyond
academic prototypes to field pilots in several countries. The practical implication is that the issue of
technology is no longer agronomic only. It's also a question of who takes risk, and who gets the value of
the risk taken.
-
STRATEGIC TECHNOLOGY CLUSTERS
The rest of the paper will review each of the six clusters of technologies, in turn. Every subsection explains
the technology in simple terms, outlines the literature on impacts, and highlights the conditions in which
the technology was applied in practice to produce the results reported. For easy reference, a summary of
representative quantitative findings is provided (Table 1).
5.1 AI-Powered Farm Management Platforms
At its core, an AI -powered farm management platform is a software solution that analyzes diverse data
streams from a farm, such as soil sensors, weather forecasts, machinery telemetry, satellite imagery, and
historical data, to provide insights and guidance on what to plant, when, and how to manage field
operations. The statistical engines used are various, ranging from supervised learning for classification
and regression, to time-series models for representation of weather and growth stage, and more recently
reinforcement learning for sequential decision making like irrigation scheduling.
Having such platforms is practical because they provide a way of aggregating a vast and ever
expanding body of agronomic knowledge into field specific advice. A farmer with 400 hectares cannot
visit every hectare themself every week. A model that identifies a proportion of the area under stress
could help focus attention and inputs where most needed, as a recent study using Unmanned Aerial
Vehicle (UAV) technology has identified stress zones over 18 per cent of the land surveyed.
The evidence on yield effects is promising but needs to be interpreted carefully. The precision agriculture
test, often referenced, demonstrated yield benefits of 20 per cent for wheat and 15 per cent for maize and
5 per cent for soya as well as a reduction in pesticide and fertilizer usage by 10 to 20 per cent, and the
sustainable yield efficiency of 92 to 95 per cent. Another study in systems that use the Internet of Things
(IoT) for monitoring and control indicates savings of 20 -30%, savings in water use of 25 -40%, and
reductions in pesticide use of 35 -50%. For an IoT data set of more than 1 million observations, LightGBM,
Decision Tree and Random Forest classifiers have reported accuracies of 98.90, 98.48 and 99.31 per cent
for classification respectively.
Table -1:Representative quantitative findings from peer -reviewed and institutional sources, by
technology cluster.Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 25
There are three cautions to be reiterated. The first one is that gains that are reported are done in relation
to a practice that is not optimal. If fertigation and variable -rate seeding are already being used, the
incremental gain will be less. The second is that the accuracy observed in the model validation is not
equal to the accuracy in the field. So, a 99 per cent accurate classifier on a test split may provide
incorrect advice during deployment in a setting with a different distribution from that of training. The third
is in operation. These platforms are continuously required to receive data and periodically need to be
calibrated. This is because the value that can be achieved in real -world testing is lower than what is
presented in lab testing in areas of low connectivity, poor sensor performance, and limited labor during
certain seasons.
A workable definition of value for these platforms is the risk -adjusted return on the next unit of decision
support, which can be expressed as in Equation (1):
where 𝛥𝜋𝑡 is the incremental profit in year 𝑡, 𝐶AI,𝑡 is the annualised cost of the platform including
subscription, hardware, and training, 𝑟 is the discount rate, 𝜆 is a risk -aversion coefficient, and 𝜎𝜋 is the
standard deviation of farm-level profits.
5.2 IoT-Based Smart Irrigation Systems
The largest use of fresh water in the world's economy is irrigation. Agriculture is responsible for about 70
per cent of total freshwater withdrawals, with up to 50 per cent of total water withdrawals being wasted
due to well managed irrigation. Smart irrigation aims to eliminate this inefficiency by using a feedback
loop to determine when and how much to irrigate based on soil moisture measurements, actual or
predicted evapotranspiration and crop growth stage.
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 26
There are three levels to represent a system. The first is a sensor layer that has capacitive or tensiometric
soil moisture probes placed at various depths, which may include other sensors measuring temperature
and humidity. The second is a decision layer, and may be as simple as a threshold rule, or may be as
complex as a machine -learning model to predict future soil moisture trajectories, based on forecast
rainfall and crop water needs. The third, generally solenoid valves located on drip or micro-sprinkler lines,
are called actuators and are told to open or close.
There is consistency between measured savings across geographies and methodologies. The US
Environmental Protection Agency (US EPA) estimates water savings of 30 per cent as a conservative
baseline figure based on an average of water savings from field and plot studies performed by the Water
Sense programme, which range from 30 to 83 per cent with a weighted average of 49 per cent. Peer -
reviewed field trials in irrigation intensive areas show savings of 30 to 55 per cent, and those in water
intensive and high-value crops, like horticultural plantations, vegetables and fruits, show higher savings of
40 to 55 per cent. In California's Central Valley, variable rate irrigation systems have been shown to
increase yields by 12-18% and reduce water use by 25 -35%, with net benefits up to 400 to 1700 dollars per
acre, amortized over the cost of the sensors.
A simple model of water applied under a sensor-driven scheme is given in Equation (2):
where ET𝑐 is the crop evapotranspiration demand, 𝑃eff is effective rainfall, 𝜃fc is field capacity, 𝜃𝑡 is the
measured soil moisture at time 𝑡, and 𝑑𝑟 is the effective root-zone depth. The practical gain over a clock-
timer system is that 𝜃𝑡 and 𝑃eff enter the decision in real time, rather than being approximated by season
average.
The risks of smart irrigation, which are often not mentioned in literature, also exist. However, false
negatives can occur due to sensor drift, biofouling, or lightning damage to sensors which may cause the
soil to look wetter than it is and result in under -irrigating. Relying on connectivity makes one point of
failure if a farm loses mobile connectivity at a crucial point. A disciplined deployment consequently
involves sensors and periodic manual checks, along with a documented fallback schedule.
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 27
Fig -2: Schematic of a sensor-driven irrigation control loop, showing soil moisture probes, a gateway with
local decision logic, cloud-based model updates, and solenoid-actuated drip lines.
5.3 Drone and Satellite Crop Monitoring
Aerial and satellite imagery convert a field into a map of indicators. The most widely used indicator is the
Normalized Difference Vegetation Index (NDVI), computed as in Equation (3):
where 𝜌NIR and 𝜌Red are the surface reflectance’s in the near -infrared and red bands respectively. NDVI
responds to chlorophyll content and canopy structure, which makes it a useful if coarse proxy for plant
vigour. Related indices such as GNDVI, NDRE, SAVI, and NDWI extend the method to green reflectance, red-
edge response, soil-adjusted canopies, and water content.
The practical difference between drone imagery and satellite imagery is one of resolution and schedule.
Drones deliver centimetre-level resolution on demand but cover limited areas per flight. Satellites deliver
metre-level or decametre-level resolution across continents on a fixed revisit cycle. Hybrid architectures
use satellites for routine scanning and drones for diagnostic follow-up on anomalies.
Published results support both platforms. A study integrating Sentinel-2 imagery with stepwise multilinear
regression and random forest classification estimated wheat grain yield prior to harvest with an 𝑅2 of 0.83
and a mapping accuracy of 86 per cent, a substantial improvement over the typical 𝑅2 of 0.5 to 0.65 for
conventional Landsat NDVI models . In sub -Saharan African smallholder systems, UAV -derived GNDVI at
five weeks after planting correlated with maize yield at 𝑟 = 0.393, outperforming hand-held SPAD readings
taken at the same stage . A UAV survey using RGB, multispectral, and thermal payloads identified crop -
stress zones over 18 per cent of the surveyed area and reduced processing time from 48 hours for
satellite-only workflows to under 4 hours . In a disease and vigour study, NDVI contributed 34 per cent of
model accuracy and healthy zones exhibited NDVI values near 0.72, consistent with expected canopy
reflectance signatures .
For the farmer, the operational product is not the index itself but a prescription. A map of NDVI zones can
be converted to a variable-rate nitrogen prescription, a scouting route for an agronomist, or a trigger for
a targeted fungicide application. The economic value depends on how precisely the recommendation is
executed by machinery and labour on the ground. A high -resolution map that cannot be followed
because the sprayer has only three sections that provides less value than a coarser map matches the
equipment available.
Two limitations should be noted. First, remote -sensing indices saturate at high biomass, which weakens
their discriminating power in dense canopies late in the season. Second, cloud cover remains a binding
constraint for optical satellites. Synthetic aperture radar (SAR) data partially addresses this problem but
has its own interpretation challenges and typically requires professional support.
5.4 Agri Supply Chain Digitisation
Agricultural value chains have faced difficulties in terms of fragmentation, lack of transparency in pricing,
post-harvest losses, and poor traceability downstream the farm gate. In many conventional systems
losses at post-harvest are estimated to be 15 -20 per cent of the gross production value. Supply -chain
digitisation aims to solve these issues by linking together three technologies: logistics and inventory ERP
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 28
software, IoT devices like GPS trackers and cold -chain temperature loggers and distributed-ledger
systems which will provide tamper-proof records of custody and condition.
Traceability, the most advanced application. A pilot by a top retailer, traced in a Hyperledger Fabric
network, cut the time to track the path of a mango shipment from about six days to a few seconds three
orders of magnitude. According to the study that incorporated blockchain along with reinforcement -
learning-based Neutrosophic multi-regression for six crops, the crop losses were predicted to be reduced
by 25 to 35 per cent in 2026 and the traceability of the supply chain was enhanced.
In addition to traceability, three other use cases are becoming popular. Input financing with smart
contracts can enable a lender to release funds upon proof of a milestone, for instance after confirmed
planting via satellite imagery, or after confirmed soil preparation via machine telemetry. Parametric crop
insurance, in which the amount is paid based on an index derived from a weather station or satellite
rather than loss adjustment has been experimented with on several platforms and offers the advantage
of reduced administration costs and quicker payouts. Lastly, farm records stored in blockchain make
credit history of smallholder farmers more verifiable, which, for the first time, will enable smallholder
farmers to access formal finance.
The restrictions are very real. Public blockchains are also limited in many ways by scalability and energy
usage, and the cost of on-chain writes can also be a challenge for smallholder transactions. The integrity
of the data entered at the first mile is the single most important determinant of the quality of a
traceability system, in which a farmer or cooperative keeps a record of what he or she has harvested
from where. With poor data in the first mile , the blockchain just immutably stores bad data. In many
deployments, issues of network access who can read and write, dispute resolution, and protection of
private data is not well resolved. The pragmatic reference design seems to be a hybrid solution of
permissioned ledgers and dissemination via QR Code and IoT data capture.
5.5 Sustainability and Carbon Tracking Tools
Carbon accounting at the farm level is more challenging scientifically and more impactful politically than
it is in most industrial sectors. Changes in soil organic carbon stocks are also spatial and temporal, as are
also the methane emissions from livestock and from paddy rice and the nitrous oxide fluxes from fertilized
fields, which are hard to measure directly. The present generation of carbon -tracking tools thus is based
on a mix of activity based models, direct measurements e.g., eddy -covariance towers or chambers, if
available and remote sensing proxies of biomass and residue cover.
A farm-level carbon balance can be summarised as in Equation (4):
where sequestration is the increase in soil organic carbon and biomass and emissions are the direct
energy consumption, embedded emissions in materials and biological fluxes of methane and nitrous
oxide in carbon dioxide equivalents. There is a large variation in the ratio between the two terms among
production systems. Among the most consistent practices that are associated with net improvements
are tillage reduction, cover cropping, and nitrogen management.
Digital tools are likely to play a role in sustainability outcomes, as evidenced through a few lines of
evidence. Pesticide and fertilizer inputs in integrated systems are 10 -20 per cent less due to precision
application of inputs, and there is a reduction in run -off, and energy use in input manufacture. Feed -
conversion efficiency is 10 -15 per cent better when using precision livestock feeding, and it also helps
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 29
lower the methane emissions per unit of product. IoT-automated processes, energy savings of up to 50
per cent have been achieved.
Policy environment is shifting rapidly. Demand for defensible farm -level data is growing because of
mandatory climate-related disclosure for large firms, as well as a growing number of voluntary carbon
markets with increasingly robust methodologies and private sector commitments to become Net Zero.
This is an opportunity to generate revenue from verified sequestration and a risk of exorbitant compliance
cost that could be an opportunity cost to productivity for small and medium farms. Design of carbon -
accounting tools is thus a question of distributive policy, and not just of science.
5.6 Farm Data Analytics Dashboards
All the above is controlled by a human interface called a dashboard. The different types of sensor
information, satellite indices, market prices, weather forecasts, and machinery telemetry are merged and
summarized in a visual manner that can be acted upon by the farm manager, agronomist, or
cooperative officer. The question of design is not in the first place a technical one. It depends on which
decisions a user takes in a typical week, and whether the dashboard directly supports the decisions.
The literature and from field experience, three principles of design come to light. First, dashboards should
display few indicators that are ready to make decisions and not be a giant matrix of raw data. Second,
they ought to reveal uncertainty in the underlying forecasts, i.e., a forecast of the yield with a broad
confidence interval should not be the same as a forecast of the yield with a narrow confidence interval.
Third, they must accommodate mobile use as this is where a lot of the relevant decision ma king is likely
to take place either in or close to the field or at an office terminal.
The economic impact reported for integrated analytics is in the range of 2 -3 years payback period on
large scale commercial farms where it is improved due to the saving of inputs, yield, and labour
reallocation. For smaller operations, the payback is more uncertain and will rely on the size of the
operations that use the dashboard and how they are paid for as a standalone subscription or as part of
other services such as agronomic advice, input supply, or insurance where the cost can be amortized
over several farms. Shared or cooperative deployments seem to be the most promising ways for access
of smallholders.
-
WHO BENEFITS, AND HOW
Numbers in one are headline impact numbers and do not necessarily reflect distribution. Who can gain
benefits and how, is at least as important as the question of whether technology works. Four groups of
stakeholders are looked at in this section.
6.1 Government and Public Sector
There are three axes where the Government benefits. One of these is food security. This boost in yield and
the potential for more stable yields, through improved information, decreases vulnerability to supply
shocks and reduces fiscal costs of emergency imports and food subsidies. The second is the matter of
water security. Smart irrigation can help postpone high cost and infrastructure investments and alleviate
aquifer stress in areas where most water is used for agriculture. The third is climate policy. A credible farm
level carbon measurement is a prerequisite for the inclusion of agriculture in a verifiable manner in NDCs.
The policy levers to be used are largely known. Improved precision tools like direct subsidies to sensor
hardware and connectivity makes precision tools accessible to farms where they would not otherwise bePartners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 30
invested. Public extension services can be the delivery channel of agronomic recommendations based
on analytics platforms and modernized with the addition of digital literacy. Data policy ownership,
portability and privacy of farm data will either contribute to value being returned to the farm or captured
completely by platform providers. The need for traceability infrastructure arises from public procurement,
which is setting buying on the condition of being verifiable production processes.
6.2 Society
Indirect but significant societal effects. The advantage is for consumers who will have access to safer
food and quicker recalls will limit the public health effects of tainted food. This decreases the amount of
pesticide in the water bodies and food, which leads to a reduction in pesticide residues and thus results in
public-health benefits. Transparent supply chain helps consumers to make informed decisions about
country of origin and ethics. Digital tools can speed up or slow down the depopulation proces s in rural
areas depending on the value they generate within the community or to those upstream. Thus, the social
issue is not so much if agriculture will be digitised or not but under what conditions.
6.3 Enterprises
The technologies discussed in this paper offer a foundation for agribusiness, food processors, and
retailers to differentiate their products on quality, place of origin and sustainability, as well as to monitor
and manage risk in long and geographically distributed supply chains. The reported three orders of
magnitude decrease in trace time is of operational interest. It transforms a food -safety event from a
multi-day crisis to a contained event. In regulated markets, supplier level carbon accounting is no longer
a reporting flourish but becomes a necessity. Businesses that purchase into integrated platforms early
will also have a better position to impact on the standards that will be set throughout the sector.
Fig -3: Mapping of expected benefits by stakeholder group across the six technology clusters examined in
Section 5.
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 31
6.4 Small and Medium Farm Operators
The small and medium operator is the one that stands to gain most from digital agriculture and is least
likely to do it at his or her own risk. So far, 68 per cent of large-scale row-crop farms have adopted yield or
soil mapping, while 13 per cent of small -scale farms had done so, USDA data indicate. The difference is
not so much a question of tastes. It's a matter of fixed costs, access to connectivity and trusted
intermediaries.
There are three delivery models that have been found to be promising possibilities for filling this gap.
Fixed costs are distributed among a wider base with cooperative deployments, where more than one
farm shares the costs of sensors, drones, and subscriptions. Bundled service offerings include analytics
within a larger commercial transaction with an input supplier, lender or buyer that lessen the cash flow
impact of the initial purchase. Public -private extension partnerships involve a government agency hi ring
a private platform to deliver extension services through local extension agents and thereby tap into the
established trust networks, helping to reduce the barriers to adopting new technology. Technology is
similar in both instances. It's a question of institutional difference.
-
IMPLEMENTATION A PRACTICAL PATHWAY
Most adoption workings stop short of that published impact figure to a lasting change on a specific farm.
There are six steps to a practical pathway that governs a gradual sequence instead of an overhaul.
The initial step is an initial audit. A farm or cooperative should record existing use and levels of losses,
water withdrawals, labour hours and yields by field and by season before any of the sensors are
purchased. There is no impact without baseline, no investment without impact.
The second step involves identifying one or two of the highest value decision(s) that is poorly informed.
Many farms will see this as irrigation scheduling, variable rate nitrogen application, or pest scouting.
Investment should be focused on technology that enables those decisions to be made, not on a generic
platform.
The third phase, a limited pilot, is conducted on 10-20 per cent of the operating area, and involves at least
one full season, and preferably two. The pilot should be such that it will yield a defensible comparison
between the treatment and the baseline, explicitly taking the weather confounds into account.
The fourth step is to interface the pilot to other the farm business. The financial accounts, insurance and
compliance reporting of the farm should be updated with information obtained from the pilot. Any
technology that only exists in a separate dashboard is likely to not survive an unfavorable season or
change of management.
The fifth Invest in Human Capital. The benefits of the use of digital tools are dependent on the levels of
agronomic knowledge of people using them. In addition to deployment, training for the farm managers,
labour and local extension officers is not a soft option. It is a core input.
The 6th step is institutionalized. As technology has shown to be beneficial in the pilot area, scaling should
involve supplier contracts, service level agreements with platform providers and operational procedures
for sensor maintenance, data backups and for fallback in the event of outages.
A simple return-on-investment framework for a farm-level deployment is captured in Equation (5):Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 32
where 𝐵𝑡 is the measured benefit in year 𝑡, including yield gains, input savings, and reduced losses, 𝐶𝑡 is
the total cost including hardware, connectivity, software, training, and maintenance, and 𝑟 is the
applicable discount rate.
-
CHALLENGES AND OPEN PROBLEMS
Five factors limit the spread of agricultural technology:
There are two main issues connectivity. Broadband internet and mobile connectivity continue to be less
uniform in both developed and developing economies in rural areas. The difference has been reduced,
although not eliminated, by low -powered, wide area networks and satellite internet. A robust adoption
plan must also be since you're not going to get connectivity all the time, and you need to plan for
gracefully falling back.
The second one is interoperability. Current ecosystem is proprietary format and closed platforms.
Farmers end up in a vendor's data environment or data integration costs are significant. The sector must
mature and open standards for sensor data, machinery, telemetry and traceability events are a
precondition.
The third one is data governance. Existing contracts vary in answering such questions as who owns the
data a farm generates, who is allowed to access it, what the terms of access are and what happens if a
provider goes out of business. Poor data governance prevents adoption and the build-up of trust.
The fourth skill is skills. This mismatch in skills is increasing in many areas between the skills needed to
work with digital farms, and the skills of the present farming population. This is especially true with the
agronomic data interpretation, where the general data analysts who don't have the agricultural
background often make wrong conclusions.
The fifth being evidence. Most of the published studies are short term season and are limited to a few
production systems. A lack of long-term trials, trials in heterodox smallholder settings, and trials looking at
the social impacts of digital uptake such as impacts on labour, tenancy, and local market structure. -
RESEARCH GAPS AND PRIORITIES
The challenges listed above result in three priorities for future research. The first are multi-seasonal, multi-
location field tests and pre -registered field designs and public data. These are costly to conduct, have
less academic side gain than methodological novelty but are the only reliable basis for policy.
The second is social and labour economics of the uptake of digitalisation. Cooperative sharing of
autonomous machinery and remote sensing leads to a reallocation of labour within a farm, within the
cooperative and within a region. What impact does it have on wages, youth staying in rural areas and on
gendered labour. Importance of these questions has not been sufficiently considered.
The third is the coupling of agricultural data and the climate and water -system models. Farms are in
watersheds and climates, and they influence these watersheds and climates. The basin -wide impacts of
return flow and recharge of the groundwater are influenced by a farm -level decision to adopt irrigationPartners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 33
technology. A better integration of the models, linking farm information with hydrological and climatic
models is necessary for better advice to the farmers and for better region policy.
- FUTURE PROSPECTS
Over the next ten years three developments can be confidently expected. One is that the basic sensing
and connectivity will continue to be reduced in cost, allowing for more small farms and lower value crops
to be viable. Second, the incorporation of the larger foundation models with agricultural data can
improve recommendation quality, especially when it comes to complicated multi -objective decisions,
like crop rotation, in an uncertain climatic context. Third, carbon and water performance will be a de facto
operational requirement in larger food value chain due to regulatory and market pressure, as described
in 5.5 the measurement infrastructure will be a de facto requirement.
A fourth development is one that remains to be seen but could be even more revolutionary. Autonomous
weeding machines, spot spraying, and other innovations using agricultural robotics have long been
touted for their potential but have had mixed results. The potential for labour, chemical and farm size
impacts will be significant if the current generation of robotics prototypes are able to achieve reliable
field performance at a competitive cost. In 10 years, this will be a transitional phase in the current debate
of precision agriculture with the combination of robotics, on -device machine learning and the data
infrastructure described here.
Fig -4: A notional roadmap of expected developments in agricultural technology, 2025 to 2035, across
sensing, analytics, autonomy, and governance layers.
-
CONCLUSION
All the evidence points to the fact that the digital transformation of agriculture isn't simply a marketing
statement, but rather a reality that's just around the corner. It is unpredictable, measurable, and
continuous and has been documented to bring yield up 15 -30 per cent in better reported studies, waterPartners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 34
saved 30 -55 per cent sensor -driven irrigation and operational cost savings typically 10 -20 per cent.
These benefits are gained but not automatic. They rely on base conditions, quality of implementation,
connectivity and skills and institutional mechanisms to determine who can capture value from data
produced at a farm. Big farms with investment and know -how have begun first and the tendency of a
digital divide between big and small is evident from the adoption data. Governments, cooperatives, and
anchor buyers all play a key role in bridging that gap, via specific subsidy, open standards, improved
extension services and procurement policies that incentivize proven practice. The research and policy
agenda over the next decade will not only be driven by the efficacy of the technologies, but also by how
to make them work for heterodox production systems, in the face of climate stress, and in a manner that
is equitable in distributing benefits. It's a big job, but there's enough evidence and the technological
platform is in place to start large scale.
REFERENCES
[1] Senapaty, M.K.; Ray, A.; Padhy , N. IoT-enabled soil nutrient analysis and crop recommendation model
for precision agriculture. Computers 2023, 12, 61.
[2] Sidik AR, Tawakal A, Sumirat GS, Narputro P. Smart Irrigation Based on Soil Moisture Sensors with
Photovoltaic Energy for Efficient Agricultural Water Management: A Systematic Literature Review.
Engineering Proceedings. 2025; 107(1):17. https://doi.org/10.3390/engproc2025107017
[3] Sivakumar, R.; Prabadevi, B.; Velvizhi, G.; Muthuraja, S.; Kathiravan, S.; Biswajita , M.; Madhumathi, A.
Internet of things and machine learning applications for smart precision agriculture. IoT Appl.
Comput. 2022, 2022, 135–165.
[4] Kolipaka, V.R.R.; Namburu, A. An automatic crop yield prediction framework designed with two -stage
classifiers: A meta-heuristic approach. Multimed. Tools Appl. 2024, 83, 28969–28992.
[5] Adetunla, A. O., Igenewari, O. O., Salau, A. O., Dada, J. O., & Kit, C. C. (2025a). Development of an
unmanned aerial vehicle for intelligent Monitoring: an insight in precision Farming. Journal Européen
Des Systèmes Automatisés, 58(5). https://doi.org/10.18280/jesa.580507
[6] Bhabad, V. M. (2025). Survey Paper on Blockchain Technology Applications in the Agricultural field.
International Journal of Advanced Research in Science Communication and Technology, 778 –785.
https://doi.org/10.48175/ijarsct-26300
[7] Sutha, K.; Indumathi, N.; Shankari, S. Recommending and Predicting Crop Yield using Smart Machine
Learning Algorithm (SMLA). Curr. Agric. Res. J. 2023, 11, 686–694.
[8] George, D. (2026b). Self -Driving Networks: AI automation for Enterprise IT. Zenodo (CERN European
Organization for Nuclear Research). https://doi.org/10.5281/zenodo.19335607
[9] Van Klompenburg, T.; Kassahun, A.; Catal, C. Crop yield prediction using machine learning: A
systematic literature review. Comput. Electron. Agric. 2020, 177, 105709.
[10] Bouni, M., Hssina, B., Douzi, K., & Douzi , S. (2024). Integrated IOT Approaches for crop recommendation
and Yield-Prediction using Machine-Learning. IoT, 5(4), 634–649. https://doi.org/10.3390/iot5040028
[11] George, D. (2026a). Real -Time IoT-Enabled smart water quality monitoring system using embedded
Multi-Parameter sensor analytics and Attention -Based predictive modeling for environmental
protection. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.18791196
[12] Development of a Sensor -Based smart irrigation Framework for sustainable agriculture. (2025).
IJIREEICE, 13(6). https://doi.org/10.17148/ijireeice.2025.13625
[13] George, D. (2026c). Seaweed -Derived Feed Additives for enteric methane mitigation in cattle:
efficacy evidence, economic incentives, and policy pathways for smallholder and commercial dairy
systems. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.22171841
[14]George, D., Dr.T.Baskar, & Srikaanth, P. B. (2026). Biometric Authentication Contemporary Trends and
Systemic Failures -A Review. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.22055699
Partners Universal Journal of Artificial Intelligence Research (PUJAIR)
Volume: 01 Issue: 04 | October-November 2026| www.pujair.com
© 2026, PUJAIR | PU Publications | DOI:10.5281/zenodo.23527989 Page | 35
[15] Hegade, M. S., & Durgha Devi, P. (2025). Blockchain – Powered traceability for safe and transparent
agriculture. International Journal of Computer Applications, 187(58), 52 –53.
https://www.ijcaonline.org/archives/volume187/number58/hegade-2025-ijca-925980.pdf
[16] George, D., George, A., & Shahul, D. (2026b). AI and IoT in Healthcare Transforming Patient Care
Through Intelligent Clinical Systems. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.18362961
[17] Mohan, R. N. V. J., Rayanoothala, P. S., & Sree, R. P. (2026). Smart agriculture 5.0: blockchain and
reinforcement learning synergy for multicropping optimization and traceable IoT -Enabled supply
chains. Frontiers in Blockchain, 9. https://doi.org/10.3389/fbloc.2026.1766232
[18]George, D., George, A., & Shahul, D. (2026a). AI and IoT in Healthcare Transforming Patient Care
Through Intelligent Clinical Systems. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.18362961
[19]Nautiyal, M., Grabow, G. L., Huffman, R. L., Miller, G. L., & Bowman, D. (2014). Residential irrigation water
use in the central Piedmont of North Carolina. II: Evaluation of smart irrigation technologies. Journal of
Irrigation and Drainage Engineering, 141(4). https://doi.org/10.1061/(asce)ir.1943 -4774.0000820
[20] Park, H. (2025). Internet of Things (IoT) for smart farming: Revolutionizing agricultural productivity
through connected technologies. International Journal of Multidisciplinary Futuristic Development,
06(02), 13–16. https://www.transdisciplinaryjournal.com/uploads/archives/20250914161238_4.pdf
[21] George, D., & Dr.T.Baskar. (2025). Securing the Future: A Review of Cutting -Edge Advances for Cloud
and IoT Cybersecurity. Zenodo (CERN European Organization for Nuclear Research).
https://doi.org/10.5281/zenodo.15288361
[22] Sanka, Y. (2026a). A review of decentralized web applications in agricultural systems. Research
Square. https://doi.org/10.21203/rs.3.rs-9450706/v1
[23] Sharma, D. a. K. (2026). AI -Powered Cybersecurity for OT and the IoT: A review of convergence, risk,
and resilience in industrial environments. Zenodo (CERN European Organization for Nuclear
Research). https://doi.org/10.5281/zenodo.20156818
[24] Vatin, N. I., Joshi, S. K., Acharya, P., Sharma, R., & Rajasekhar, N. (2024). Precision Agriculture and
Sustainable Yields: Insights from IoT -Driven Farming and the Precision Agriculture Test. BIO Web of
Conferences, 86, 01091. https://doi.org/10.1051/bioconf/20248601091
[25] Wahab, I., Hall, O., & Jirström, M. (2018). Remote sensing of yields: Application of UAV Imagery -Derived
NDVI for estimating maize vigor and yields in complex farming systems in Sub -Saharan Africa.
Drones, 2(3), 28. https://doi.org/10.3390/drones2030028
[26] Xing, Y., Liu, X., & Wang, X. (2026). Integrating UAVs, satellite remote sensing, and machine learning in
precision agriculture: pathways to sustainable food production, resource efficiency, and scalable
innovation. Frontiers in Agronomy, 7. https://doi.org/10.3389/fagro.2025.1670380
论文原文与出版信息
Zenodo (CERN European Organization for Nuclear Research)