FlowsBook a callSign inCreate free accountSign up
Workflows by industry

How grain buyers check recent satellite images of farm fields

Get a recent satellite picture of every field you buy from, every few days, with the date and cloud cover, and know what a 10 m picture can and cannot show.

Part of the guide Workflow automation

By the xMap team · · 5 min read

Grain buyers can look at recent satellite images of farm fields without visiting them: list the fields you buy from with a point for each, and every few days pull the latest clear picture of each one, with the date it was taken and how cloudy it was. At about 10 metres a pixel, a picture shows whether a field is green, golden, harvested, bare or partly under water. It does not count bushels.

This article is for grain merchants, cooperatives, millers and maltsters who contract crops from many farms and want an early look at how the season is going in each field. It explains what the pictures show, how clouds get in the way, a made-up example from a buyer near Rosario in Argentina, and how to have the pictures arrive by email.

What a 10 metre picture shows, and what it does not

The pictures come from the Sentinel-2 satellites, in natural colour, roughly as the eye would see the field from above. Each pixel covers about 10 by 10 metres. A field of 40 hectares is about 630 metres across, so it fills a picture with roughly 60 pixels a side: plenty to see its colour and any large patch that looks different.

You can usually seeYou cannot see
Green crop, ripening crop, harvested stubble, bare soilYield, moisture content or grain quality
Large flooded or waterlogged patchesSmall wet spots under a few metres
Whether harvest has started across the fieldWhich variety is planted
Burned or very damaged areas of some sizeDisease or pests before they change the colour of a large area
Changes between one picture and the nextAnything under cloud on the day

A rough guide to natural-colour pictures at about 10 m a pixel.

These are natural-colour pictures only. They are not a vegetation index or a yield model, so treat them as a quick look that tells you where to ask questions, not as a measurement.

Clouds, and why a clear picture is not guaranteed

A satellite passes over a given place every few days. On many of those days there is cloud. You set a limit on cloud cover, for example 20%, and pictures cloudier than that are skipped. Two things to know:

  • The limit is for the whole scene, a large square many kilometres wide. A picture rated 10% cloud can still have one cloud right over your field; one rated 40% can have your field perfectly clear. If you keep missing pictures, raise the limit and look at what comes back.
  • Wet seasons are hard. In a cloudy month you may get no clear picture at all. Widen the period from the last 7 days to the last 14 or 30, and the newest picture under the limit is used. Each field comes back with satellite_taken empty when none was clear enough, so you know which fields were not seen rather than seeing an old picture without realising.

A made-up example near Rosario

Say a grain merchant near Rosario contracts soybeans and maize from 45 farms in the surrounding countryside, and wants to see how close each field is to harvest before sending trucks and agreeing delivery dates.

  1. The merchant keeps a table of 45 fields: farm name, crop, contract number and a point in the middle of each field.
  2. Every Monday and Thursday a flow looks up the latest picture of each field from the last 14 days, under 30% cloud.
  3. Each field gets the date of its picture, its cloud cover, how many clear pictures there were in the period and a link to the picture itself.
  4. The buyer receives one email with a row per field and its picture. Fields with no clear picture are listed first, so someone can phone those farms.

Looking down the email on a Thursday, the buyer sees a group of fields to the north already turning from green to gold and three fields to the south with dark patches that were not there on Monday. The first group goes to the top of the truck plan; the farmers in the second get a call about standing water. All of this is invented to show how the steps fit together.

Making the pictures useful to the team

  • Put the point in the middle of the field, not at the farmhouse. Set how far the picture reaches; 1,500 m from the point shows about 3 km across, which may take in neighbours' fields, so a smaller distance suits small plots.
  • Keep a history. Pick Every one in the period on a single field and you get one row per clear picture, up to 10, to compare across a month.
  • Use the larger size for detail. 1024 px pictures are sharper in the email; 512 px are quicker and enough for a glance.
  • Mind the limits per run. Up to 50 fields are looked at in one run and up to 25 pictures are made. For more fields, split the list across two flows or two days.

Building it as an AI employee

  1. Schedule: twice a week, early morning.
  2. My data: your table of fields with their points.
  3. Satellite images: Each incoming place, Last 14 days, most cloud cover 30%, Get the picture on.
  4. Sort and limit: fields with no clear picture first.
  5. Email one per run to the buying desk, or Show on a map so the team can click each field.

The weekly field watch from space template follows this shape, and weekly satellite picture of a site is the single-point version for one large farm. Before signing a new supplier, land offers, checked from space and the ground adds a picture to what you already know about the land. For more on the pictures themselves, see satellite imagery.

Questions

How often is there a new picture?

The satellites pass every few days, but cloud means a clear picture can take longer. Checking twice a week with a 14-day window usually gives a recent one.

Can it estimate yield from the picture?

No. The pictures are natural colour at about 10 m a pixel. They show the field's state, such as green, ripe, harvested or flooded, not how much it will produce.

Why did a field get no picture?

Every picture in the period was cloudier than your limit. Raise the limit or widen the period, and look again.

How many fields can one flow check?

Up to 50 fields a run, with up to 25 pictures made. Larger lists can be split across flows or days.

Templates that do this

Read next

More on workflows by industry · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.