What a digital twin of a farm looks like

A digital twin of a farm is a computer model of the real farm, its fields, soils, crops, weather and machines, that is kept up to date with data from the farm itself. It differs from a map or a set of records because it changes as the farm changes, and because you can ask it questions about what is happening now and what might happen next.

The term comes from engineering, where a twin mirrors a machine or a factory. Our general guide to what a digital twin is covers the idea. This one looks at what it means on a farm.

Three things that make it a twin

Plenty of farm software shows fields on a map. Three features separate a twin from a picture.

1. It is tied to real, identified things. Each field, soil zone, sensor, machine and store in the model corresponds to one in the real world and has a stable identity over time.
2. It stays current. New data flows in, from weather feeds, sensors, machines and records, so the model reflects the state of the farm today and keeps its history.
3. It can be questioned. You can ask what the state of a field is, how it got there, and what is likely to happen under different conditions or choices.

A twin does not need to be elaborate or three-dimensional. A well-organised model of thirty fields with sound data is more useful than an impressive graphic built on guesswork.

The layers of a farm twin

A farm twin is best thought of as layers that share the same geography.

The land. Field boundaries are the foundation. Everything else attaches to them. Alongside boundaries sit elevation, slope and aspect, drainage, watercourses, hedges, tracks and buffer zones.

The soil. Soil type and texture, organic matter, pH and nutrient results from sampling, and where available soil scans that show variation within a field. This layer changes slowly.

The weather. Past, present and forecast conditions for each field: temperature, rainfall, wind, humidity. Values derived from weather belong here too, such as growing degree days, frost risk and spray windows. Weather at field level is not the same as at the nearest station, as our guide to microclimate explains.

The crop. What is growing where, the variety, sowing date, growth stage, and the rotation history of the field. Satellite and drone imagery add a view of crop vigour and show patchiness within a field. Yield maps from the combine close the loop at harvest.

Water. Soil moisture, from sensors or from water balance estimates, plus irrigation systems and their records where they exist.

Operations. What was done, where, when, with what and by whom: cultivation, drilling, applications, harvest. Modern machines record much of this as they work. The rest comes from job sheets and farm records.

Assets and stores. Machinery, its location and service state, grain stores, inputs in the shed, and for mixed farms the livestock, housing and grazing.

Paperwork. Assurance schemes, environmental agreements, application records and contracts. These are tied to fields and dates, so they fit naturally onto the same model.

What you can ask a twin

The purpose of joining layers together is to answer questions that no single layer can.

• What is the state of things now? Which fields reached a given growth stage this week? Which are too wet to travel on?
• Why did that happen? A poor patch in a yield map may line up with a soil zone, a drainage problem, an old field boundary or a missed operation. With the layers together you can check each one.
• What happens if? How does the harvest schedule look if the next fortnight is wet? What does a change of rotation do to workload in the autumn?
• What needs doing, and what needs recording? Which jobs are due, and are the records for the completed ones in place?

The first two are about knowing. The third is forecasting and simulation. It depends on models, which are simplifications and carry uncertainty. A twin can set out the likely consequences of a choice. It does not replace the judgement of the farmer, the advice of an agronomist, or the requirements on a product label.

Why it is harder than it sounds

Most of the work in a farm twin is not modelling. It is getting the data to agree with itself.

• Identity. The same field may be "Top Forty" in the farm office, a number in a government land register, a different name in the machinery terminal, and half of a larger block in last year's records. Working out that these are the same thing is a task called entity resolution.
• Boundaries change. Fields are split, merged and re-fenced. The twin needs to keep history without losing track.
• Formats differ. Machines, sensors, laboratories and satellites all deliver data in their own formats and at their own intervals.
• Gaps and errors. Sensors fail, rain gauges block, and someone forgets to log a job. A twin should show where data is missing or estimated, not hide it.
• Provenance. For every value you should be able to see where it came from and when. That is what makes an answer checkable.
• Ownership and access. Farm data has commercial value. Be clear about who holds it, who can see it, and what happens to it if you change supplier.

A sensible way to start

You do not need every layer at once.

1. Get the field boundaries and names right, and settle on one identifier for each field.
2. Add field-level weather, which is useful on its own from the first day.
3. Attach the records you already keep: cropping, operations and soil tests.
4. Add sensors, imagery and machine data where they answer a question you actually have.
5. Decide which questions matter most and check that the twin can answer them.

Each step is worth having even if you never take the next one.

Where Prism fits

Prism describes its work in four verbs: See, Understand, Plan and Act. See means connecting records, sensors and machines into a digital twin. Understand means asking a question and getting an answer that shows its source. In agriculture, Prism Microclimate, built with FarmGrid, covers 961k fields across one whole state with field-level weather, soil moisture, frost, spray and field-work windows, growing degree days and a morning brief per field. A fuller farm twin of the kind described here is what we would build next on that base. There is more on our agriculture page.