Observing a working farm: a reference design

01What a farm needs to see

Observing a working farm is mostly about water. Agriculture is the largest user of withdrawn water: at global level, the withdrawal ratios are 69 percent agricultural, 12 percent municipal and 19 percent industrial².

Two measurements decide when to irrigate. One is how much water the weather is taking from the crop. The method recommended by the Food and Agriculture Organization computes reference evapotranspiration from weather data, so values from different places and seasons can be compared¹.

The other is how much water the soil still holds where the roots are. Soil probes at several depths measure it directly, and the international guide to instruments and methods of observation gives soil moisture a chapter of its own³.

A view from above completes the picture. A publicly accessible database built from satellite data monitors agricultural water productivity in near real time, and lets the farm compare its fields with the area around them⁴.

02Siting probes, mast and weather station

Probes go where the decisions are made. Each irrigation zone gets at least one, placed in soil and crop that represent the zone rather than at its wettest or driest corner. Where soils change within a field, the probe follows the soil, not the field boundary.

The weather station stands over short grass, away from the irrigated plots and from buildings. The standard method defines reference evapotranspiration for a hypothetical grass reference surface, so the station should measure the weather over something like it¹.

The gateway sits on a tall mast at the edge of the holding, high enough to see every probe over the crop. Probes sit close to the ground, so the mast does almost all of the work of keeping the radio path clear.

The microgrid sits beside the mast, so the gateway and the farm edge share one power supply and one place to visit.

Probes and the weather station run on their own batteries, so they can sit anywhere in the field without cables. Every probe position is surveyed and recorded, with its depths and the soil around it, so that its readings can still be read correctly after the crop changes.

03From sensor to decision

The probes, the weather station and the flow meters send short reports by long-range, low-power radio to the gateway on the mast. The industry alliance that maintains the protocol publishes its radio parameters for each regulatory region worldwide⁵.

The gateway passes everything to the farm edge, a small computer beside it. There the weather data becomes reference evapotranspiration, the probe readings become a soil water balance per zone, and the two together become an irrigation schedule.

Satellite water maps arrive when the link allows, and they are used to check the probes against the wider field rather than to replace them⁴.

Every value carries its time, probe, depth, calibration and quality flag, and leaves the farm through an open interface for sensor data, so an adviser or water agency can read it without a special integration⁶.

04Shared spectrum is the constraint

Long-range farm radios use licence-exempt spectrum that is shared with many other devices, so each transmitter may only use a small share of the time. In Europe, the harmonised standard limits the band from 868.0 to 868.6 MHz to a duty cycle of 1 percent unless the device uses polite spectrum access⁷.

Time on air depends on the spreading factor and the payload. Slower settings reach farther but stay on the air longer, so they use up the allowance faster and drain the battery sooner.

This design keeps reports short and infrequent, which suits soil moisture because it changes over hours, not seconds. The simulator shows the trade between spreading factor, payload size and the uplinks per hour the duty-cycle limit allows.

05When the links fail

Farms are often at the end of the line, and the internet link is the part most likely to fail. In this design the irrigation schedule runs on the farm edge, so losing the link stops reports to advisers but not water to the crop.

While the link is down, the edge keeps every value and replays the record in order, with original timestamps, when the link returns. Satellite maps simply arrive late.

If the gateway fails, the probes keep measuring but cannot report. The edge then falls back to a weather-only schedule from the standard method, and flags each zone as estimated until the probes are heard again¹.

A probe that fails or drifts is shown as a gap with its reason, never filled in silently, so no one irrigates a zone on a reading that is not there.

Power failures are handled the same way. The microgrid carries the gateway and the edge through the night, and when its battery runs low the edge stops forwarding to advisers before it stops listening to the probes.