Variable-Rate Nitrogen from Satellite Maps: A UK Field Guide

Most UK arable farms now have free satellite maps of every field, often inside software they already pay for. Far fewer turn those maps into a nitrogen plan that saves money or adds yield. The step between a colourful NDVI picture and a spreader that changes its rate across the field is where most of the value is won or lost.

This guide covers one job: using variable rate nitrogen satellite imagery to set the main spring nitrogen dressing on winter cereals, mostly wheat, with some notes for barley and oilseed rape. If you want the wider picture first (drones, soil scanning, yield mapping and how they fit together), start with our overview of satellite, drone and variable-rate agronomy and come back here when you’re ready to write an actual prescription.

What the satellite actually measures

The free imagery behind nearly every UK platform comes from the European Space Agency’s Sentinel-2 satellites. Each pixel covers 10 m by 10 m, and a satellite passes over a given field roughly every five days. Because UK springs are cloudy, you might get a usable clear image only every two or three weeks in March and April.

The satellite records reflected light, including near-infrared, which healthy green leaf reflects strongly. Software turns that into an index:

• NDVI (Normalised Difference Vegetation Index) is the classic one. It works well on thin early crops but “saturates” once the canopy closes, so a good crop and a very good crop look the same.
• NDRE and other red-edge indices hold their contrast better in thicker crops, which makes them more useful from GS31 onwards.
• GAI (Green Area Index) is what an agronomist actually wants. Some platforms estimate it directly from the imagery. AHDB’s canopy management guidance is built around GAI, which is why it matters.

The “AI” part of most of these tools is less dramatic than the marketing suggests. Machine learning models convert raw reflectance into a GAI or biomass estimate, mask out cloud and shadow, and group pixels into management zones. They’re useful, but they don’t know why a patch is thin, and that question decides whether it should get more nitrogen or less.

The tools UK farmers are using

You probably don’t need a new subscription. Check what you already have before buying anything.

Prices change every season, so treat these as ballpark figures. Free tiers usually give you the imagery and a basic zone map. Paid zone or prescription services commonly run from roughly £2 to £8 per hectare per application, and agronomy-bundled services fold the cost into the advice fee. A mounted N-Sensor costs tens of thousands of pounds new, so on most farms it’s a contractor job.

Should thin areas get more nitrogen or less?

You need to settle this before you open any software, because the software will happily do either.

More nitrogen to thin areas (“building the canopy”) is the logic behind early-season VRN. A patch with a GAI of 1.0 at GS30 needs help to reach the roughly 2.0 to 3.0 GAI that AHDB guidance links with good yield potential in wheat. A patch already at 3.5 is at risk of lodging and can be cut back.

Less nitrogen to thin areas makes sense when the patch is thin for a reason nitrogen won’t fix: compaction on old headlands, a waterlogged hollow, blackgrass, take-all, a stony bank that always burns off in June. Putting extra N there pays for fertiliser the crop can’t turn into grain.

The satellite can’t tell these apart. You can, with a walk across the field. In practice most UK growers do this:

1. Use the satellite map to find the variation.
2. Visit the thin and the thick zones with a spade and a quadrat.
3. Override any zone where the cause is structural, and leave “more N to thin” for zones that are simply behind (late drilling, slug damage, poor establishment on a clay knoll).

Yield maps from previous seasons help a lot here. A zone that has been thin in the satellite image and low in the yield map for three years is a soil problem, not a nitrogen shortage.

Worked example: 24 ha of winter wheat in Lincolnshire

Here’s a realistic field that we’ll take all the way through.

The field. 24 ha of Group 1 milling wheat after oilseed rape, medium soil, SNS Index 1. The RB209 total is around 220 kg N/ha before any milling top-up. The farm is in an NVZ.

The plan before VRN. Three splits:

• Late February to early March: 40 kg N/ha flat (tillering)
• Early April, GS30 to GS31: 120 kg N/ha (the main dressing, and the one we’ll vary)
• Late April to early May, GS32: 60 kg N/ha flat, plus a later milling dose

The imagery. The first clear Sentinel-2 image after the first split arrives on 28 March. The platform’s GAI layer shows a range from 0.9 to 3.4 across the field. Its automatic clustering proposes four zones.

Ground truthing. A 40-minute walk on 30 March finds:

• Zone A (low GAI) on the south-east corner: late-drilled after a wet spell, thin but healthy plants with no structural problem.
• Zone B (low GAI) along the old field boundary: compacted and wet, with a history of low yield maps.
• Zones C and D: normal to forward crop, with D showing heavy, lush growth on a former muck heap site.

The prescription. The aim is to keep the total close to 120 kg/ha × 24 ha = 2,880 kg N, while moving nitrogen to where it will work.

Zone B goes against the “more to thin” rule on purpose, because the walk showed the crop couldn’t use extra N there. The prescription ends up 95 kg of N lighter than a flat 120.

The money. Ammonium nitrate at 34.5% N bought at, say, £350/t works out at about £1.01 per kg of N. The 95 kg saved is worth roughly £96 across the field, or £4/ha. That alone won’t pay for much.

The real gain is yield and quality. With wheat at £180/t, grain is worth £0.18/kg, which gives a breakeven ratio of about 5.6 (the figure you’d look up in RB209’s tables). If Zone A responds with an extra 0.4 t/ha because it caught up, that’s 1.4 t, or £252. If Zone D avoids lodging, which could easily cost 1 t/ha plus a slow, wet harvest, the saving on 5 ha is several hundred pounds more.

A fair expectation for a field like this is £10 to £30/ha of net benefit in an average season. In some seasons the benefit is close to zero. Independent UK trials have been mixed: many show modest gains, and some show no difference from a well-judged flat rate. Most of the value comes from avoiding mistakes in the extreme zones, not from fine-tuning the middle.

Getting the prescription into the spreader

Plenty of farms have the map but never manage to apply it. Check these points in the order you’ll meet them.

Spreader capability. You need a spreader with electronic rate control that accepts a prescription, either through the ISOBUS task controller or through the spreader’s own controller. Amazone ZA-TS, Kuhn Axis, Bogballe M-line, Kverneland Exacta and Rauch Axis models with ISOBUS all handle this, usually with an activation code for task control. Older mechanical spreaders can’t.

File format. Most terminals accept ISO-XML (TASKDATA folder) or a shapefile. Export from the platform, put it on a USB stick, and load it before you leave the yard. Here’s what a shapefile’s attribute table looks like for the example above, which is a good thing to open and check before trusting the export:

ZONE_ID  AREA_HA  RATE_KG_N  PRODUCT          RATE_KG_PRODUCT
A        3.50     150        AN 34.5%N        435
B        2.50     90         AN 34.5%N        261
C        13.00    120        AN 34.5%N        348
D        5.00     95         AN 34.5%N        275

Watch for one common mistake. Some terminals want the rate in kg of product, and others want kg of nitrogen. 120 kg N/ha of AN is 348 kg of product. Getting this wrong means you apply a third of the intended N, or nearly three times as much.

Zone size versus spread width. A 10 m pixel means nothing to a 36 m spreader. Set the platform to smooth zones to at least the bout width, or set a minimum zone size of about 0.5 ha. Otherwise the controller chases tiny changes it can’t physically deliver, and the spread pattern blurs them anyway.

Rate change lag. Spinning-disc spreaders take a few seconds to change rate. At 12 km/h, that’s 10 to 15 m of travel. Many terminals have a “look-ahead” setting, so set it according to the spreader manual.

Caps. Put a ceiling and a floor on every zone. A typical guard is no zone more than 30% above or below the planned flat rate. This stops an odd cloud-shadow pixel turning into a 200 kg/ha strip.

Cloud, timing and what to do when the sky won’t clear

April in the UK regularly goes two or three weeks without a clean Sentinel-2 pass. You have three options.

• Use the most recent clear image if it’s less than about 10 to 14 days old and the crop hasn’t moved through a growth stage. Relative differences between zones tend to hold even if absolute GAI has risen.
• Use a paid high-frequency source. Some platforms buy in commercial imagery (Planet’s daily 3 m data is the usual one), which raises the chance of a clear day. Expect to pay more per hectare.
• Switch to a tractor-mounted sensor or drone. A contractor with a Yara N-Sensor reads the canopy live and doesn’t care about cloud. A drone flight with a multispectral camera (for example a DJI Mavic 3M) gives a centimetre-level map on demand, although processing takes longer than most people expect.

Don’t use an image from before the first split to set the second. The crop will have changed, and you’ll be correcting for a picture that no longer exists.

Records, NVZ rules and SFI

Variable-rate application doesn’t change your legal limits. In an NVZ, the N max for winter wheat is 220 kg N/ha averaged across the crop area on the holding, with an allowance of 20 kg/ha for each tonne of expected yield above 8 t/ha. VRN can push individual zones above the field average, which is allowed as long as the farm-level crop average stays within N max and every rate is justified by your nutrient management plan.

Keep the as-applied file from the terminal, not just the plan. It’s your evidence of what actually went on, and it’s the file an inspector or a Red Tractor auditor can check against your fertiliser records. Most platforms will store it automatically if the terminal syncs. Otherwise copy it from the USB stick into your record-keeping software (Gatekeeper, Muddy Boots, Farmplan or similar) the same day.

If you’re in an SFI agreement with a nutrient management action, the satellite maps, zone notes and as-applied files are good supporting evidence for the nutrient management review. A one-page note per field (“Zone B held back due to compaction, see photo log”) is worth more in an audit than the map alone.

A one-hour routine for the main dressing

When the first clear image after the first split lands:

1. Open the GAI or NDRE layer for every wheat field (5 minutes).
2. Pick the fields with a GAI range wider than about 1.0. Fields with less variation than that are rarely worth a variable-rate plan, so spread them flat.
3. Walk the low and high zones on those fields with a quadrat and a spade (30 to 40 minutes per two or three fields).
4. Build the prescription, set the caps, and check whether the terminal expects product or N (10 minutes).
5. Export, load, and run a short test strip near the gateway to confirm the rate changes before doing the whole field.

If something looks wrong, such as a zone boundary running dead straight along a tramline or a “thin” patch that turns out to be a cloud shadow, trust the walk over the map. The imagery is good at showing where to look. Deciding what each zone gets is still your job.