Satellite crop insurance and lending data for each plot

How satellite data supports farm loan underwriting

An agricultural loan file usually rests on what the applicant declares: where the land is, what they'll grow and how much of it. Satellite data lets you test each of those statements before money goes out, which is the part of agricultural credit risk assessment that used to need a field visit.

Does the field exist where the applicant says?

The plot boundary is laid over current imagery. A boundary that lands on a road, a river, buildings or forest is an obvious flag. One that overlaps another applicant's plot is less obvious and worth checking across the whole book.

Is the crop the one declared?

Crops have recognizable growth curves. A vegetation index such as NDVI rises as the canopy closes and falls at ripening and harvest, and the timing and shape differ between, say, rice, maize and sugarcane. Comparing the season's curve with the declared crop and sowing date shows whether they fit.

Is the area right, and how has the plot done before?

We measure the acres inside the boundary and compare them with the declared area. Imagery from past seasons shows whether the field was cropped each year, left fallow, flooded or damaged, and that history, together with weather and field condition, goes into the plot's risk profile.

Monitoring a loan or insurance portfolio in season

Once loans or policies are out, the question becomes "which part of my book is in trouble?" Nobody reviews thousands of plots one at a time, so work by exception: compare each plot's crop health and yield estimate with what's normal for that crop and region at that point in the season, and look closely at the plots and districts that fall well short.

• Regional totals of expected yield show where a drought or flood is hitting the book, weeks before repayments come due.
• Plots that stop developing, or show no crop at all after the declared sowing date, are worth a call from the loan officer.
• Plots with high pest-risk scores can get guidance through AI farm advisory while the problem is still small.

What you do with an early warning (reschedule, visit, send advice or wait) is still a credit decision for your team.

Claims verification and area-yield checks

For satellite crop insurance, claims are where the season-long record is most useful. A claim says something happened to a crop, on a field, around a date, and each part can be checked. Imagery from before and after the event shows whether crop health dropped and where. Radar helps after floods and storms, when cloud often hides the field from optical satellites. The field's own yield estimate and nearby fields growing the same crop put the claimed loss in context.

1. Each claim, filed against a plot ID that matches a boundary on file, is compared with the satellite record for that plot and its neighbors.
2. Claims that fit the record move ahead under your settlement rules.
3. Claims where the area, crop or timing don't fit go to an adjuster for a field visit.

Some losses don't show from above. Hail damage to fruit or a drop in grain quality may leave the canopy looking normal, so those claims still need someone on the ground. Our article on how satellite imagery is used in crop insurance covers more examples.

Index and parametric insurance: where satellite data fits

Traditional indemnity cover pays for a loss assessed on each farm. Index insurance pays when a measured value crosses a threshold agreed in advance, such as rainfall at a weather station, the average yield across an area, or a vegetation index over a region. Parametric is the wider term for covers that pay on a trigger instead of an assessed loss. Payouts can be quick and cheap to administer.

Their weak point is basis risk, the gap between what the index measures and what happened in a particular field. A farmer can lose a crop and get nothing, or get paid when the crop was fine. Satellite data helps at three stages. Field- and region-level histories of vegetation, moisture and evapotranspiration let a product team test how a trigger would have behaved in past seasons. During the season, an independent record shows whether conditions in the insured area match what the index measured. After a payout, comparing the insured fields' actual condition with the payout shows where basis risk sits. Our piece on parametric crop insurance and AI loss adjustment goes further into product design.

What data you provide and how integration works

The checks are only as good as the plot records behind them. For a pilot or a full rollout, we'd ask for:

If all you hold is a single GPS point for each farm, the crop and area checks get much weaker, so it pays to collect boundaries at application. Through the agriculture data API, one call per field (POST /submitField with your API key) returns every satellite view, the values behind it and the field report as JSON, which your credit or claims system can store against the plot ID.

Model risk and ground-truthing

Satellite checks can be wrong. These are the usual weak spots:

• Small plots. At 10-meter resolution a half-acre plot covers about 20 pixels, and the ones along the edge mix the crop with paths, bunds or the neighbor's field.
• Intercropping and mixed plantings, which blur the growth curve of any single crop.
• Long cloudy spells. Radar fills many gaps, but optical indices like NDVI can still have holes in the record.
• Wrong boundaries, which produce confident results for the wrong piece of land.

The fix is regular ground-truthing. Each season, visit a random sample of plots, including some the data says are fine, and compare what your staff see with what the record showed. Over a few seasons that shows how much weight satellite flags can carry in your crops and regions. Treat a flag as a reason to look, and keep the decision with a person.

Regulation and fairness, in general terms

Bringing new data into a credit or claims decision carries obligations that depend on your country and license. Treat this as a list to raise with your compliance team rather than legal advice.

• Plot boundaries tied to a named farmer are personal data under laws such as the EU's GDPR, so tell applicants what you collect and why.
• If satellite data contributes to a declined loan or a rejected claim, your staff should be able to explain what the data showed in terms the farmer understands.
• Check that satellite-based rules don't consistently disadvantage farmers with very small plots, where the data is weaker, or regions with frequent cloud.
• Give farmers who disagree with what the imagery shows a way to ask for a field visit.

In the EU, the AI Act (Regulation (EU) 2024/1689) lists AI systems used to assess the creditworthiness of individuals as high-risk, which brings extra requirements for the lender using them.