IoT in Agriculture: How Connected Farming Actually Works
Agricultural IoT is not one network. A single farm can run LoRaWAN, NB-IoT, LTE-M, Cat 1 bis, 4G, 5G, GNSS and satellite at once, because each solves a different job. The skill is not connecting the farm. It is choosing the right connection for each task.
IoT in agriculture is the use of connected sensors, machines and software to measure, monitor or control farm operations. There is no single farm network. Low-power radios such as LoRaWAN and NB-IoT carry tiny readings from soil, livestock and remote infrastructure; 4G, 5G and Cat 1 bis handle richer data and machinery; GNSS adds position; edge computing decides what happens locally; and satellite reaches where terrestrial coverage stops. The network follows the application, never the other way round.
Agriculture is about measurement, not gadgets
Agriculture is one of the clearest illustrations of why the Internet of Things is not really about things connected to the internet. It is about measurement. A grower wants to know whether a field needs water before irrigating it, whether a cow has stopped moving normally before anyone notices it is ill, whether a grain store is turning humid, whether a remote trough is empty, or whether a pump has failed three miles from the farmhouse.
Sometimes that means a sensor sending twenty bytes an hour. Sometimes it means a camera, an autonomous vehicle or a drone moving gigabytes. That gap of several orders of magnitude is what makes agricultural connectivity interesting. There is no single farm network, because no single technology serves both ends of that range well.
UK policy increasingly treats this as core to how farms will operate. The England Farming Roadmap 2050, published by Defra on 24 June 2026, sets out a 25-year plan that leans heavily on technology, innovation and better farm data. It came with an extra £53 million for the Farming Innovation Programme, taking total innovation funding this year to £123 million and including dedicated funding rounds for robotics. A separate £20 million agri-tech robotics and automation competition opened in August 2026, part of a commitment to invest at least £200 million in agricultural innovation by 2030. The direction of travel is clear: automation, robotics and data are being treated as the future of the sector, and connectivity is the layer underneath all of it.
What can actually be connected on a farm
Almost anything that can be measured, located, switched, counted or controlled. Soil probes can report moisture, temperature, conductivity, pH and nutrient indicators; weather stations add rainfall, humidity, wind and frost risk. Instead of treating a field as one uniform block, measurements reveal the differences across it, which is the basis of precision agriculture: applying water or fertiliser where and when it is actually needed. The sensor does not make the decision. It provides the data from which a grower, agronomist or control system can make a better one.
Irrigation: where monitoring becomes control
Water is the clearest example of IoT moving past a dashboard. Rather than irrigating on a fixed timetable, a system can combine soil moisture, rainfall, temperature, forecast and crop need, decide whether irrigation is actually required, and then operate valves or pumps itself. It becomes a feedback loop: measure, communicate, analyse, decide, actuate, then measure again. The LoRa Alliance lists irrigation control among the major agricultural applications for LoRaWAN, drawing on soil moisture, rainfall, water levels and weather.
Livestock joins the network
The same logic applies to animals. A collar or ear tag can track location, movement, activity, temperature, grazing behaviour, geofence boundaries and health indicators. A cow cannot tell a farmer its behaviour has changed since yesterday. A sensor can, which means an abnormal pattern is flagged earlier rather than replacing the stockperson's judgement. The GSMA has used livestock tracking as an example NB-IoT application, and LoRaWAN is widely used for location, welfare and behaviour. It also shows why agricultural IoT is not simply putting a SIM in everything: a herd of 1,000 animals does not need 1,000 conventional 4G connections when each device sends only a trickle of data.
The connectivity toolkit, job by job
The useful way to think about agricultural connectivity is as a ladder of increasing communications requirement, not a contest for a single winner. Each technology earns its place on a specific kind of task.
LoRaWAN: connecting the field
A small battery-powered sensor can reach a gateway over kilometres using very little energy. A farm installs its own gateway, connects many devices to it, and backhauls to the internet over 4G, 5G, Ethernet or satellite. The key distinction is between local device connectivity and internet backhaul: a soil sensor does not need a SIM card, it needs to reach a gateway, and the gateway reaches the internet. That lets hundreds or thousands of small devices share a handful of connections. LoRaWAN passed 125 million deployed devices globally by the end of 2025, growing at a 25 percent compound annual rate, with agriculture named as one of the environments driving that scale.
NB-IoT: when the sensor talks to the network directly
LoRaWAN is not always the answer. Sometimes a device is too isolated to justify a local gateway: a water tank several miles out, a lone weather station, a remote pump, or a sensor on separately worked land. Where NB-IoT coverage exists, the device connects straight through the mobile network. NB-IoT was designed for small amounts of data, low device complexity and low power, which suits kit that spends most of its life asleep: wake, measure, connect, send a tiny packet, sleep again. It is well suited to greenhouse environmental monitoring and remote fixed sensors.
LTE-M: the mobile middle ground
LTE-M was also designed for IoT, but supports higher data rates and mobility more naturally than NB-IoT. That makes it attractive where a device needs more than a tiny static sensor but less than full broadband cellular: livestock trackers, movable equipment, asset tracking, machinery telemetry and devices needing richer firmware updates. The honest question is never NB-IoT versus LTE-M. It is what the device is doing.
LTE Cat 1 bis: normal LTE, designed sensibly for IoT
Cat 1 bis fills a genuinely useful gap. It provides conventional LTE connectivity without the complexity of higher-category devices and, unlike traditional Cat 1, can operate with a single receive antenna. That matters for compact trackers, controllers, monitoring equipment and sensor hubs. It offers far more standard IP connectivity than constrained LPWAN radios while staying appropriate for lower-cost hardware. Cavli Wireless, for instance, positions its C-series modules, including Cat 1 bis variants with integrated GNSS and optional eSIM, for precision-farming tasks such as irrigation, pest detection and predictive machinery maintenance; its published smart-agritech case study uses a single C-series module to combine cellular, satellite positioning and over-the-air updates in one package. The trend is important: not every low-cost agricultural device needs an LPWAN radio.
4G: still the workhorse
Not every application needs a specialist network. For many farm buildings, ordinary 4G is exactly right. A single industrial 4G router can connect an entire installation of PLCs, controllers, CCTV, refrigeration, grain monitoring, pumps and energy meters, providing Ethernet, Wi-Fi, VPN, firewalling, remote management, serial and Modbus connectivity and cellular failover. The router becomes the site's WAN connection. That is a completely different architecture from fitting every sensor with its own cellular link, and it illustrates a general IoT principle: do not use a wide-area radio where a local network will do.
5G: powerful, but not everywhere
Putting a 5G modem in a soil sensor would be absurd. Putting 5G into an autonomous machine makes far more sense, because the data requirements differ by orders of magnitude. An advanced agricultural vehicle can carry multiple cameras, radar, LiDAR, GNSS, inertial navigation, controllers, obstacle detection and edge AI. Teltonika Networks describes exactly this with its RUTM50 router: mounted in the cab or on the roof, it connects over Ethernet to the tractor's main controller, which in turn links to hydraulic and electric actuators, soil and weather sensors, GNSS and INS units, radar, LiDAR, ultrasonic sensors and a camera. The vehicle builds a model of its environment, plans a safe path and operates autonomously, with Wi-Fi allowing remote operation if something fails. Now bandwidth and latency matter.
ZTE and China Mobile used standalone 5G to automate rice production on around 12,000 acres of marginal land near Da'an City, Jilin Province. Remote-controlled machinery, smart irrigation and drones let one worker monitor and control several machines at once from a control room rather than driving a single vehicle. That is a fundamentally different use of connectivity from reporting soil moisture once an hour, and both are agricultural IoT.
RedCap: bringing 5G down to IoT size
Full 5G is unnecessary and expensive for many devices. Reduced Capability, or RedCap, trims device complexity while keeping the useful parts of the 5G ecosystem, and agriculture is emerging as a real application. It is not for a soil sensor, but for a machine, controller, camera or industrial device where NB-IoT is too constrained and full 5G is overkill.
A GSMA case study from the Waigang farm in Jiading, Shanghai, pairs a 5G private network with 5G RedCap terminals across 40 pieces of machinery, including tractors, rice transplanters, plant-protection drones and harvesters. Ploughing, planting and harvesting were made unmanned across 1,600 acres of rice fields, one operator controlled multiple machines from a control room, and labour costs fell by 53 percent. RedCap kept the terminal cost and power low enough to make the economics work.
NTN and satellite: reaching where networks do not
Agriculture has an obvious connectivity problem. Farming happens where the land is; mobile networks are built where population and economic activity justify the infrastructure. Those maps do not always line up. That makes agriculture a natural fit for Non-Terrestrial Networks. 3GPP Release 17 introduced support for NB-IoT and eMTC over satellite, extending cellular IoT skyward. Increasingly the question is not which mobile network covers this field, but whether the device can see the sky. UK operators are already moving in this direction, as our analysis of O2 Satellite for M2M sets out.
Satellite does not mean every sensor carries a satellite modem. The architecture is usually a sensor reaching a gateway over LoRaWAN, with the gateway using satellite only for expensive long-distance backhaul. That lets a farm blanket a large area with low-cost, low-power LoRaWAN devices while paying for satellite on just one link.
The Banalytics project in Ghana and Brazil, supported by Lacuna Space, uses satellite-connected LoRaWAN sensors to catch Black Sigatoka, the most damaging banana disease, before it spreads. Around ten instrumented plants per hectare track temperature and humidity, soil-nutrient sensors sit on a 50-metre grid, and AI imaging flags early onset, with data from one monitored hectare informing the wider growing area. It runs where conventional networks do not reach, and it helps growers cut unnecessary fungicide.
The connected farm, drawn out
Put the technologies on one mixed farm and the picture is not a single futuristic network. It is a collection of small, practical improvements, each connected by whatever radio suits the job. The diagram below shows a realistic mix converging on the same edge and cloud layer.
| Application | Typical connectivity | Why |
|---|---|---|
| Soil moisture sensor | LoRaWAN / NB-IoT | Tiny data, years on a battery |
| Weather station | LoRaWAN / NB-IoT / Cat 1 bis | Small data, often fixed and remote |
| Livestock tracker | LoRaWAN / LTE-M / NB-IoT | Low data, but moving and power-limited |
| Remote water tank | LoRaWAN / NB-IoT | Isolated, infrequent readings |
| Irrigation controller | LoRaWAN / Cat 1 bis / 4G | Two-way control, must stay reachable |
| Farm office / buildings | Fibre / 4G / 5G | General site WAN via one router |
| CCTV | Ethernet or Wi-Fi locally, 4G/5G backhaul | Continuous video, local first |
| Tractor telemetry | Cat 1 bis / LTE-M / 4G | Moderate data on the move |
| Autonomous machinery | 5G / private 5G | High bandwidth, low latency, safety-critical |
| Field beyond cellular coverage | LoRaWAN plus satellite / NTN | Local radio, satellite backhaul only |
| Machinery positioning | GNSS / RTK | Centimetre accuracy for guidance |
The network follows the application. The application is never distorted to fit the network. Get that order right and the technology choices mostly make themselves.
Beyond the radio: position, edge and the cloud
Connectivity tells you what happened. GNSS tells you where. That matters enormously in agriculture, for tractor guidance, asset and livestock tracking, geofencing, drone navigation and precision spraying. Where ordinary GNSS is not accurate enough, RTK positioning can deliver centimetre-level precision under the right conditions, letting machinery follow highly repeatable paths and apply inputs to specific locations rather than roughly across a field.
Put a camera on a tractor and a new problem appears. A camera generates far more data than a soil sensor, and streaming every frame to the cloud may be unnecessary, costly or impossible. An edge computer on the machine can process imagery locally: detect a weed, record its location, trigger the sprayer, and send only the event, location, confidence and a sample image upward. That cuts bandwidth and keeps the machine working when the link drops. In agriculture, intermittent connectivity should never stop an operational process. A tractor must not become helpless because it drove behind a hill.
Edge processing does not make the cloud redundant. It changes what goes there. The cloud handles long-term storage, fleet management, dashboards, analytics, maintenance history, firmware and integration with wider farm software. The sensible split is usually real-time decisions locally, fleet-wide intelligence centrally. And once data reaches the internet, something still has to route and interpret it, which is where MQTT and APIs quietly do the work: a sensor publishes a reading, an irrigation controller and a dashboard subscribe, and an analytics platform combines it with a rainfall forecast. A successful system is not sensor to internet. It is measurement, transport, interpretation, decision, action.
Getting it right in the field
The radio is only part of the design. Several practical factors decide whether an agricultural deployment survives contact with an actual farm.
Power and antennas
A device in the middle of a field needs electricity, whether from batteries, solar, machinery power, mains or energy harvesting. This is why low-power technologies matter: if a battery costs £5 but sending someone across the estate to change it costs £100, battery life is an economic decision, not a specification-sheet curiosity. The antenna is forgotten even more often. A sophisticated router on a poor antenna inside a steel cabinet can perform worse than a modest modem on a correctly installed external antenna. Metal machinery, steel buildings, low-mounted cabinets, vegetation, long cable runs and terrain all degrade signal, so antenna position, height, frequency support and cable loss belong in the design, not the retrofit. Our guide to RF interference and resilience covers the same physics in more depth.
Security: a farm is now an OT network
A hacked soil sensor is not frightening. A remotely controllable irrigation valve, pump, grain-ventilation system, refrigeration unit, autonomous vehicle or access system is another matter. Connected agricultural equipment needs the same discipline as industrial IoT: unique credentials, no default passwords, encrypted communications, secure remote management, VPNs, firewall rules, least-privilege access, firmware updates, a device inventory and network segmentation. Above all, do not expose farm equipment directly to the public internet just because remote access is convenient. Convenience is not a security architecture.
Much of this comes down to how devices are addressed and reached remotely, which is where the difference between private and public IP on cellular IoT becomes a security decision rather than a networking detail.
Resilience and network choice
Many agricultural systems run unattended. If home broadband fails, someone notices; if a remote irrigation controller stops communicating on a Friday evening, nobody may find out until Monday. Useful safeguards include watchdogs, ping-and-reboot monitoring, dual or multi-network SIMs, store-and-forward data, local automation, battery backup and cellular-plus-satellite fallback. The rule is that loss of internet must not mean loss of local control: a sensible controller keeps running safe local rules even if its cloud platform disappears for a while.
Agriculture is also a strong argument for multi-network connectivity. An operator's coverage map can look excellent while one particular field stays problematic, because terrain does not negotiate. Depending on the application, a deployment might use a single-network SIM, roaming or multi-network connectivity, dual SIM, eSIM or satellite fallback. For thousands of scattered devices, remote SIM provisioning matters: over a ten-year deployment, being able to change operator, tariff or roaming arrangement without visiting 2,000 remote sensors is a real operational advantage. That is the practical value of IoT eSIM and SGP.32, and it is why they matter more in agriculture than they first appear. The value is not better signal. It is operational flexibility over a very long life.
How to choose a technology for a farm
Do not start with whether to use LoRaWAN or NB-IoT. Start with the application, and let the answers narrow the field:
- What are you measuring or controlling? A reading, or a process you can act on?
- How much data does it produce? A temperature is tiny. Video is not.
- How often must it communicate? Once a day and ten times a second are different worlds.
- Does it move? A cow, a tractor and a fixed tank have different needs.
- Is mains power available? If not, energy use becomes fundamental.
- What happens when communications fail? Can it store data and keep operating safely?
- What coverage exists at the actual installation point, not on a coverage map?
- How long will the equipment stay deployed? Hardware often outlives the network chosen for it.
- How many devices will there be? Ten cheap SIMs may not matter. Ten thousand do.
Work through those and the shortlist usually resolves itself. The farm of the future is not entirely futuristic: an autonomous tractor on RTK and 5G may sit fifty metres from a cheap soil sensor on LoRaWAN, while an old pump still speaks Modbus over RS-485 through a gateway on 4G, and an isolated station falls back to satellite. None of those technologies has won. They are doing different jobs. Real-world programmes bear this out: Australia's Agriculture Victoria On-Farm IoT Trial, which began in 2020 and cost about A$12 million, rolled out a LoRaWAN network across roughly 32,000 square kilometres and awarded around 350 grants to test the technology across dairy, horticulture, cropping and sheep. The measure of agricultural IoT is not the number of connected devices. It is whether anyone on the farm would miss the system if it disappeared tomorrow.
