In short: A modern greenhouse monitoring system measures light, temperature, humidity, CO2, substrate moisture and EC on a few-minute cycle, and hands the readings to a climate computer that adjusts vents, screens, heat and irrigation. Under UK glass the sensors are wireless, the building is wet steel, and the crop grows into the radio path every summer.
Key Takeaways
- The glasshouse is an industrial control system — Priva and Hoogendoorn climate computers make thousands of decisions a day from sensor data, so a dropout is not a missing chart, it is a control loop running blind.
- The radio environment changes with the crop — A tomato plant reaching four metres by July is a column of water in every propagation path, which is why a network surveyed in February fails in high summer.
- Robots arrive next, and they need more than sensors did — Autonomous spray, scouting and picking platforms from Saga Robotics, Dogtooth and others need continuous coverage down every aisle, not the patchy Wi-Fi that a data logger tolerates.
In a nutshell
What is a greenhouse monitoring system, and what does it actually control?
A greenhouse monitoring system is the sensor layer feeding a climate computer. In a commercial UK glasshouse that means measuring air temperature and humidity at several heights, photosynthetically active radiation inside and outside, CO2 concentration, pipe and screen temperatures, substrate water content and electrical conductivity in the rockwool slabs, drain volume and drain EC, plus wind speed and direction from the weather mast. The climate computer, usually a Priva, Hoogendoorn or Ridder system, turns that into vent positions, screen positions, heating pipe temperature, CO2 dosing, fogging and irrigation start times.
The scale is easy to underestimate. Thanet Earth in Kent, the country's largest glasshouse complex, covers roughly 90 hectares of site with its glasshouses running combined heat and power engines that supply electricity, heat and the CO2 that goes into the crop. Low Carbon Farming built large tomato glasshouses in Norfolk and Suffolk heated with recovered warmth from Anglian Water treatment works. The Lea Valley Growers Association still represents a dense cluster of cucumber and pepper growers across Hertfordshire and Essex. These are process plants with plants in them, and they are managed by a very small number of people per hectare.
Growers care about this because the margins are thin and the physics is unforgiving. UK protected horticulture has had a hard decade: gas prices in 2022 pushed a meaningful share of the glasshouse area to stay unplanted the following season, and the empty supermarket shelves of February 2023 were the visible end of that decision. What is left is a sector that competes with the Netherlands and Spain on yield per square metre, and yield per square metre is a control problem.
Why do wireless sensors drop out under glass?
Because a glasshouse is a moving, wet, metallic structure that changes shape twice a day and changes character twice a year.
The frame is steel: gutters every eight metres or so, trusses, and hanging gutter systems that carry the crop on steel rails. Thermal screens close at dusk and open at dawn, and an aluminised energy screen is a large reflective sheet drawn across the entire roof of the compartment, which changes propagation across the whole house within a couple of minutes, twice a day, every day. Fogging and misting systems put water in the air deliberately. Condensation runs on everything at dawn.
Then there is the crop. A tomato or cucumber crop planted at knee height in December reaches three to four metres by midsummer, and the plants are mostly water at a spacing designed to intercept every available photon. That is close to the worst case for 2.4 and 5 GHz absorption, and it arrives gradually, so the failure mode is a system that worked perfectly in spring and starts logging gaps in July. Growers frequently blame the sensors and replace batteries. The batteries were usually fine.
Add the annual reset. Between crops the house is stripped, the floor and structure are washed down and disinfected, and anything mounted at crop height comes out or gets soaked. Whatever survives goes back in a slightly different place. A network that depends on precise access point placement does not survive that cycle well.
Polytunnels are a different set of problems with the same outcome. Soft fruit sites in Kent, Herefordshire and Angus run tens of hectares of tunnels that are moved between fields, on holdings where the nearest fixed line is the farm office half a mile away and the mobile signal is whatever the nearest macro site offers. Tunnel skins go on and come off seasonally. There is nothing stable to build a fixed network around.
What does a gap in the data actually cost?
Three things, in ascending order of pain.
The first is control quality. A climate computer running on a stale reading will hold a vent or a screen in the wrong position, and in a house where a few degrees of humidity deficit separates healthy transpiration from Botrytis conditions, that matters within an hour. Growers who have lost sensor feeds describe it as flying with one instrument out: nothing dramatic happens immediately, and then the disease scout finds something.
The second is irrigation. Substrate sensors such as the Grodan GroSens family drive water content and EC strategy in rockwool, and a mid-afternoon dropout in a high-radiation week means either over-irrigating and flushing nutrient to drain, or under-irrigating and stressing a crop that is carrying fruit. Both cost yield, and the water and fertiliser going to drain is also a nutrient management problem where the site sits in a sensitive catchment.
The third is traceability. Retailer audits under LEAF Marque and the Red Tractor protected crops standard, plus increasingly detailed residue and water-use reporting, all assume the grower can produce continuous records. The same argument that applies to the cold store applies here: a gap in the record is a finding, whatever actually happened in the crop.
What changes when the robots arrive?
Protected cropping is where UK agricultural robotics is closest to being ordinary. Saga Robotics has run Thorvald platforms delivering UV-C light at night to suppress powdery mildew in strawberry tunnels. Dogtooth Technologies has been picking and grading strawberries with autonomous robots on commercial UK farms for several seasons. Fieldwork Robotics came out of Plymouth chasing raspberries. Glasshouse operators are trialling autonomous scouting trolleys that photograph every plant on every row for pest and disease detection and truss counting.
A data logger tolerates an intermittent connection because it caches and resends. A robot does not. Autonomous platforms need continuous coverage down every aisle for teleoperation handover, safety supervision and image offload, and image offload in particular is an uplink problem: a scouting run over a hectare produces tens of gigabytes of imagery that must reach a server before the next run starts. Uplink is the one thing consumer Wi-Fi and public mobile in rural Kent or rural Norfolk are least likely to give you.
This is where a private network earns its place. An Ofcom Shared Access licence in the 3.8 to 4.2 GHz band, from £80 a year at low power, gives the grower licensed spectrum across the site. Radios can be planted on the gutter line and re-planned per crop cycle without pulling fibre through a house that gets pressure-washed every year. One network then carries the climate sensors, the substrate probes, the robots, the packhouse handhelds, the CCTV on the intake gate and the site's staff accommodation Wi-Fi backhaul, which on a site running seasonal workers is not a small consideration.
Does the same network help with energy and CO2?
It does, and this is the part that tends to move a board. A glasshouse with CHP is a power station with a horticultural attachment. The engines generate electricity, the heat goes into the buffer tanks and the pipe rails, and the cleaned exhaust CO2 goes to the crop. Whether to run the engines at a given hour is a live trading decision against wholesale power prices, and the same asset can be dispatched into balancing services.
Making that call well needs the heat demand forecast, the buffer tank state, the CO2 demand of the crop and the actual climate data in one place, updating continuously. Growers doing this by hand off a screen in the office are leaving money on the table, and the ones automating it need the underlying telemetry to be reliable enough to trust with an automated dispatch decision. The same sensor grid that keeps the crop healthy is the input to the energy decision.
Where should a grower start?
Survey during the peak crop, not in the empty house. The February coverage map is the optimistic one and it is the one most integrators produce.
Then pick the compartment that has caused the most trouble, instrument it properly on a licensed network, and run one full crop cycle against a compartment on the existing setup. Compare the dropout counts, not the yield: the yield has too many variables to attribute cleanly in one season, but the completeness of the data record is a clean measurement and it is the thing the robots and the auditors both depend on.
If the site is already planning a robotics trial for next season, do the network first. A trial that spends its season debugging connectivity produces no useful evidence about the robot, and the supplier will blame the signal either way.